It has only a slight connection to Firefox, but I did meet people there doing database-ish things with Firefox extensions to provide a whole experience....
Freebase (http://www.freebase.com) is an odd service. It is a wiki-like database. Anybody can go in, look at a table and decide there needs to be another column on it and add a new column. Anyone can put new data in. Anyone can create new "types" (like tables). There are APIs for managing time information and geo information and all that kind of stuff. They have created the concept of a "Base" which sort of combines a social situation with a custom view of data. It allows a better separation between the model and the view.
The tools are mostly Javascript or Google-app based and they work pretty well.
And how many places can one go to and talk about the re-contextualization of socially-constructed semi-structured database tables into dynamic entities and not have people get glassy-eyed and start looking for open doors or windows to jump out of?
Sunday, November 09, 2008
Sunday, June 22, 2008
Mozilla Dev Kit ... and Scratchbox?
Seeing a post on the Mozilla Developer Kit (MozDevKit? MDK?) has gotten me very psyched. I have been trying to add some things to the developer doc site. I see, for example, a lot of really good information in blog posts and then the corresponding DevMo pages are ... not so up-to-date. In general, there is much that can be done to make developing Mozilla more accessible to people and MDC does not get as much love as it deserves from people who are doing development and should therefore be able to write. But they probably have no time to doc because they are, you know, coding. It is a conundrum.
But build environments are complicated. How can setting up such an environment be made easy?
Can a Mozilla development kit can be set up in a trackable, verifiable, chroot-ed environment using a tool such as scratchbox? I have been trying to get a xulrunner (FF Mobile) build working to an embedded device using scratchbox and it has been challenging, to say the least, but interesting. Tools like scratchbox or crosstool seem a very good idea. And every production build environment, such as the one I used while I was in the Dev Tools group at Apple, uses a verifiable, chroot-ed environment. It is really hard to get reproducibility without it, and without reproducibility, it is really hard to help anybody who has problems. This is why most responses to build issues on the newsgroups consist of 'please set up X, Y, and Z, and do A, B and C, and see if that helps....'
Also, the usual use of scratchbox is to build on host A for a different architecture, target B. But A and B can actually be the same. MoCo builds Linux binaries on Linux and Mac OS X binaries on Mac OS X, but that does not have to be so. Of course, scratchbox is very Linux-centric. And I see that a Linux version of the MDK is the first item listed on the aforementioned blog post. This is probably true because Linux is amenable to scripted, verifiable building. But it would be interesting (perhaps for only me) to consider how a scratchbox on Linux could build Mac OS X binaries, or how a scratchbox on Mac OS X could build binaries for either Linux or the Mac. And of course, some might be interested in making Windows development possible, but that is way beyond my bailiwick. Yo no hablo Microsoft and I am just fine with that. Can you imagine, for example, getting the Microsoft development tools into a scratchbox environment? I cannot, and I am sure it would be ugly, but it might be possible.)
Downloading a single file, launching it and then finding oneself in a known, trackable, verifiable virtual host for building Mozilla apps seems a compelling opportunity.
But build environments are complicated. How can setting up such an environment be made easy?
Can a Mozilla development kit can be set up in a trackable, verifiable, chroot-ed environment using a tool such as scratchbox? I have been trying to get a xulrunner (FF Mobile) build working to an embedded device using scratchbox and it has been challenging, to say the least, but interesting. Tools like scratchbox or crosstool seem a very good idea. And every production build environment, such as the one I used while I was in the Dev Tools group at Apple, uses a verifiable, chroot-ed environment. It is really hard to get reproducibility without it, and without reproducibility, it is really hard to help anybody who has problems. This is why most responses to build issues on the newsgroups consist of 'please set up X, Y, and Z, and do A, B and C, and see if that helps....'
Also, the usual use of scratchbox is to build on host A for a different architecture, target B. But A and B can actually be the same. MoCo builds Linux binaries on Linux and Mac OS X binaries on Mac OS X, but that does not have to be so. Of course, scratchbox is very Linux-centric. And I see that a Linux version of the MDK is the first item listed on the aforementioned blog post. This is probably true because Linux is amenable to scripted, verifiable building. But it would be interesting (perhaps for only me) to consider how a scratchbox on Linux could build Mac OS X binaries, or how a scratchbox on Mac OS X could build binaries for either Linux or the Mac. And of course, some might be interested in making Windows development possible, but that is way beyond my bailiwick. Yo no hablo Microsoft and I am just fine with that. Can you imagine, for example, getting the Microsoft development tools into a scratchbox environment? I cannot, and I am sure it would be ugly, but it might be possible.)
Downloading a single file, launching it and then finding oneself in a known, trackable, verifiable virtual host for building Mozilla apps seems a compelling opportunity.
Friday, May 16, 2008
the age of bugs, and other useless information...
In a recent post, dbaron suggested that people inappropriately judge Mozilla by, for example, the age of the bugs that one can see. I think he is right, but I think there is a reason for this. It is usually completely impossible to look at a bug and figure out whether it will be fixed or not. There are long-standing reasons for the use of nobody as an assignee, but this can be un-helpful.
Mozilla code development is socially driven. This may seem obvious, but the effects of this are not necessarily obvious, especially to someone in coming from commercial software. Look at the recent revisions of the module ownership system and you can see how tenuous the ownership of code can be. Nobody has to care about any bug. Or, to say it another way, only 'nobody' actually has to care about the vast majority of bugs.
It is very hard to look at a bug and see who actually cares about it. No. I will not gripe about bugzilla. Really. But I will just say that anything in a bug is not what it says, but how that is looked at by which group does the work. Priority, Confirmed, and many other terms mean different things to different groups of people. And everyone seems to be ok with that.
I have filed many bugs thinking that Firefox should just 'do the right thing' and that people would want to know. Is still have this idealism at times, but I recognize the other viewpoint as being valid. It may be true that one should not fix something that does not need to be fixed.
And as dbaron has pointed out to me many times, one is always welcome to file a patch. And, actually, that turns out to be the key to many things. I used to think there was value in asking about a bug to understand it. I used to think there was value in discussing a solution before creating a patch. But, as a practical matter, the number of people willing to talk about making changes is vastly more huge than the number of people who will make changes. So, considered questions almost always get silence while patches that are crap get responses. If I have 10+ years of database experience and ask a question about sqlite use by Places, and a 12-year-old asks a question with a patch he came up with in five minutes, who will get a response? Who will be more likely to get something started on a bug? It seems obvious to me now.
So when in doubt, submit a patch. Any patch.
So, this leaves a question. How does Mozilla want to manage bugs submitted by or read by or tracked by non-developers? It seems clear that bugzilla is really best only for developers. Asking users to look at bugzilla and make sense of what they see is, I would suggest, not realistic. Maybe the idea behind Hendrix needs to be developed further. A lot of what is in bugzilla looks like graffiti in a train station. So maybe something is needed manages that graffiti and relates it to actual bugs that people may actually work on. But then, the people who do the work do not mind looking at the graffiti and just working on what they talk about. They seem to know which ones are actually significant, so why fix it if it's not broken?
Mozilla code development is socially driven. This may seem obvious, but the effects of this are not necessarily obvious, especially to someone in coming from commercial software. Look at the recent revisions of the module ownership system and you can see how tenuous the ownership of code can be. Nobody has to care about any bug. Or, to say it another way, only 'nobody' actually has to care about the vast majority of bugs.
It is very hard to look at a bug and see who actually cares about it. No. I will not gripe about bugzilla. Really. But I will just say that anything in a bug is not what it says, but how that is looked at by which group does the work. Priority, Confirmed, and many other terms mean different things to different groups of people. And everyone seems to be ok with that.
I have filed many bugs thinking that Firefox should just 'do the right thing' and that people would want to know. Is still have this idealism at times, but I recognize the other viewpoint as being valid. It may be true that one should not fix something that does not need to be fixed.
And as dbaron has pointed out to me many times, one is always welcome to file a patch. And, actually, that turns out to be the key to many things. I used to think there was value in asking about a bug to understand it. I used to think there was value in discussing a solution before creating a patch. But, as a practical matter, the number of people willing to talk about making changes is vastly more huge than the number of people who will make changes. So, considered questions almost always get silence while patches that are crap get responses. If I have 10+ years of database experience and ask a question about sqlite use by Places, and a 12-year-old asks a question with a patch he came up with in five minutes, who will get a response? Who will be more likely to get something started on a bug? It seems obvious to me now.
So when in doubt, submit a patch. Any patch.
So, this leaves a question. How does Mozilla want to manage bugs submitted by or read by or tracked by non-developers? It seems clear that bugzilla is really best only for developers. Asking users to look at bugzilla and make sense of what they see is, I would suggest, not realistic. Maybe the idea behind Hendrix needs to be developed further. A lot of what is in bugzilla looks like graffiti in a train station. So maybe something is needed manages that graffiti and relates it to actual bugs that people may actually work on. But then, the people who do the work do not mind looking at the graffiti and just working on what they talk about. They seem to know which ones are actually significant, so why fix it if it's not broken?
Sunday, March 30, 2008
A Really Neat Test Reusability Situation
I had something happen yesterday which makes me think there might be hope. For my brain and how I sometimes fail to keep track of things. I was asked, in a bug I filed a while back, if the failure was reproducible.
I do not why it was not reproducible earlier. I hardly ever find out why people do not see the bugs I see. I filed the bug and it got the usual responses: "works for me", "what are the running? are you sure?", "i see something weird", "why are we checking this?", "can you attach more information?", "did you configure the thingamagig to build the whatamacallit?", "can you attach a lot more information?". It just gets really ... exhausting. Actually, this particular bug was not as bad as some others. And it is nice when someone can actually reproduce something I filed.
But right now, if you look in my ~/mo directory, you see 3 different builds of the browser, a half-dozen nightlies (some a few months old), a couple of calendar builds, a few xulrunners, an nspr build (what the heck was I doing with that?), a re-usable profile directory, and other random stuff. So, do I still see my bug? I look in the directory and and it is, again, exhausting, even to just think about.
Then I saw something.
tests_20080106_160400/
You know, I was talking about a re-usable test product, wasn't I? Oh, yeah.
% cd tests_20080106_160400
% ls
dist/ runReftests* testing/ xpcshell-simple/
%
% ./runReftests ~/mo/browser_20080327_235255/dist/MinefieldDebug.app/Contents/MacOS/firefox ~/mo/debug.mozilla_20080106_173138_PST
Holy test failures, Batman! It worked! I just gave it a fairly recent build and that re-usable profile directory I mentioned and there it is. Lots of failures.
Now, I am thinking again of that standalone test product I was mentioning. Once again, the buzzing in my head is beginning. There is apparently a new way to build a product out of the Mozilla build system. Create a directory, put some stuff in it, and voila! A new product. Now, to find the documentation for this wonderful new capability. Ok, I am sure that people are busy. Is it fair to ask them to document things? Should I ask a question in the newsgroups? I will just get that response that I really hate so much: "what exactly are you trying to do?". I think I was trying to ask a question. Again, it can just be exhausting. I even created an MDC page. Seeing a place for something like this to to be documented, someone who knows what is going on might, you know, document something. Not even a nibble on that yet. O well. You can lead them to water....
And, finally, I have to do homework today. My Stats and PDE homework is not doing itself. And why don't I get more Mozilla stuff done?
I do not why it was not reproducible earlier. I hardly ever find out why people do not see the bugs I see. I filed the bug and it got the usual responses: "works for me", "what are the running? are you sure?", "i see something weird", "why are we checking this?", "can you attach more information?", "did you configure the thingamagig to build the whatamacallit?", "can you attach a lot more information?". It just gets really ... exhausting. Actually, this particular bug was not as bad as some others. And it is nice when someone can actually reproduce something I filed.
But right now, if you look in my ~/mo directory, you see 3 different builds of the browser, a half-dozen nightlies (some a few months old), a couple of calendar builds, a few xulrunners, an nspr build (what the heck was I doing with that?), a re-usable profile directory, and other random stuff. So, do I still see my bug? I look in the directory and and it is, again, exhausting, even to just think about.
Then I saw something.
tests_20080106_160400/
You know, I was talking about a re-usable test product, wasn't I? Oh, yeah.
% cd tests_20080106_160400
% ls
dist/ runReftests* testing/ xpcshell-simple/
%
% ./runReftests ~/mo/browser_20080327_235255/dist/MinefieldDebug.app/Contents/MacOS/firefox ~/mo/debug.mozilla_20080106_173138_PST
Holy test failures, Batman! It worked! I just gave it a fairly recent build and that re-usable profile directory I mentioned and there it is. Lots of failures.
Now, I am thinking again of that standalone test product I was mentioning. Once again, the buzzing in my head is beginning. There is apparently a new way to build a product out of the Mozilla build system. Create a directory, put some stuff in it, and voila! A new product. Now, to find the documentation for this wonderful new capability. Ok, I am sure that people are busy. Is it fair to ask them to document things? Should I ask a question in the newsgroups? I will just get that response that I really hate so much: "what exactly are you trying to do?". I think I was trying to ask a question. Again, it can just be exhausting. I even created an MDC page. Seeing a place for something like this to to be documented, someone who knows what is going on might, you know, document something. Not even a nibble on that yet. O well. You can lead them to water....
And, finally, I have to do homework today. My Stats and PDE homework is not doing itself. And why don't I get more Mozilla stuff done?
Saturday, March 29, 2008
Participating in Earth Hour?
Are you participating in Earth Hour? I am, apparently. My wife says that we can and should turn off all lights in the house. We can then turn off all electrical devices (even my Server machine!) and, further, she is convinced we will survive and may perhaps even escape psychological damage.
I am dubious.
If you do not hear back from me, you know why. Good luck.
I am dubious.
If you do not hear back from me, you know why. Good luck.
Monday, January 07, 2008
A Standalone Test Product
What does one need to do to help test Firefox. It turns out that to help very much, it takes a fair amount of effort. Just downloading the latest browser and "doing stuff" with it does provide testing, but it is not very complete. That kind of testing tends to be shallow at best.
Firefox has great automated testing tools built into it, and so one can build a copy of Firefox using '--enable-tests' as a configuration flag. One gets a bunch of standalone executable tests, some xpcshell tests, the mochitest functionality and reftest. There is a lot there to work with.
However, building Firefox can be a pretty hefty requirement. Is it really necessary? One benefit of automated testing is that the more of it that gets done, the more valuable it can be. There are lots of OS versions and machine configurations to worry about. MoCo can either try to have a copy of all those OSes and all those machines, or it can allow the community to meaningfully assist with testing without jumping through all the extra hoops needed to be able to build the browser. After all, the software required to just run tests is much simpler.
In pursuit of this, I am creating a standalone product which can be a test vehicle for testing Firefox. After all, if Firefox is changing every day, should one have to refetch all those tests? One would hope that when Firefox is changing a lot, the tests are not changing, so that the tests can provide confidence for the changes being made.
So far, I have a standalone product that I can use, on my Mac OS X system, to run reftests. There are issues with some of the other technologies that I will try to address. For now, I have taken the last 4 nightly builds and used the same test product on them and got the same result.
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/bidi-000.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/bidi-001.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/bidi-001-j.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/mixedChartype-02.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/mixedChartype-02-j.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/line-breaking/currency-2.html
There may already be bugs on these test failures. It is hard to tell. I cannot seem to find the obvious way to locate bugs filed against automated tests.
The test product is based on the contents of the "
There are issues, already, that make it a little difficult to just grab this test product and put it up in a shareable location for other Mac OS X users. For one thing, script bits need to be put inside the application bundle of the app being tested. For reftest, it is only three files that need to be put inside the ".app/Contents" directory, but it is awkward to do manually.
Looking at mochitests, I can already see problems. The tests all assume that they can do something like "../../../dist/bin/testExecutable". There is no reason they need to assume that, but they all do. Does the xpcshell executable that runs the tests need to be the same build as the Firefox browser itself? I would say no, but that is the way it has always been done. Until issues like these are teased apart, there will have to be lots of workarounds.
If anyone has any suggestions, I am open to them. If MoCo wants to make automated testing available to a much larger group of people, without massively increasing their app download times, it seems that a standalone test product would be one way to do it.
Firefox has great automated testing tools built into it, and so one can build a copy of Firefox using '--enable-tests' as a configuration flag. One gets a bunch of standalone executable tests, some xpcshell tests, the mochitest functionality and reftest. There is a lot there to work with.
However, building Firefox can be a pretty hefty requirement. Is it really necessary? One benefit of automated testing is that the more of it that gets done, the more valuable it can be. There are lots of OS versions and machine configurations to worry about. MoCo can either try to have a copy of all those OSes and all those machines, or it can allow the community to meaningfully assist with testing without jumping through all the extra hoops needed to be able to build the browser. After all, the software required to just run tests is much simpler.
In pursuit of this, I am creating a standalone product which can be a test vehicle for testing Firefox. After all, if Firefox is changing every day, should one have to refetch all those tests? One would hope that when Firefox is changing a lot, the tests are not changing, so that the tests can provide confidence for the changes being made.
So far, I have a standalone product that I can use, on my Mac OS X system, to run reftests. There are issues with some of the other technologies that I will try to address. For now, I have taken the last 4 nightly builds and used the same test product on them and got the same result.
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/bidi-000.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/bidi-001.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/bidi-001-j.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/mixedChartype-02.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/bidi/mixedChartype-02-j.html
- REFTEST UNEXPECTED FAIL: file:///.../testing/reftest/layout/line-breaking/currency-2.html
There may already be bugs on these test failures. It is hard to tell. I cannot seem to find the obvious way to locate bugs filed against automated tests.
The test product is based on the contents of the "
_tests" directory as built with the '--enable-tests' flag, with some minor changes. Also, the contents of the directory are actually files, not links back into a source tree. The basic directory structure right now looks as so:
tests\
Minefield.app/Contents/MacOS/chrome/reftest.jar
Minefield.app/Contents/MacOS/chrome/reftest.manifest
Minefield.app/Contents/MacOS/components/reftest-cmdline.js
testing\
mochitest\
reftest\
xpcshell-simple\
There are issues, already, that make it a little difficult to just grab this test product and put it up in a shareable location for other Mac OS X users. For one thing, script bits need to be put inside the application bundle of the app being tested. For reftest, it is only three files that need to be put inside the ".app/Contents" directory, but it is awkward to do manually.
Looking at mochitests, I can already see problems. The tests all assume that they can do something like "../../../dist/bin/testExecutable". There is no reason they need to assume that, but they all do. Does the xpcshell executable that runs the tests need to be the same build as the Firefox browser itself? I would say no, but that is the way it has always been done. Until issues like these are teased apart, there will have to be lots of workarounds.
If anyone has any suggestions, I am open to them. If MoCo wants to make automated testing available to a much larger group of people, without massively increasing their app download times, it seems that a standalone test product would be one way to do it.
Wednesday, December 05, 2007
A Systematic Look at Testing, Perhaps
Seeing the release of Firefox 2.0.0.10, followed by the rapid release of 2.0.0.11, made me wonder about how regression testing is going at Mozilla. Of course, there are no simple answers to be found for any of the issues. We all want less bugs and for bugs to get fixed more quickly. Nor am I feeling that there is some deep and dark hidden flaw that is coming to the fore here. Things could be better. No shock there. Things could be worse. Lots of people care enough to make sure they are not. And MoCo moves from one day to the next. We all love the products that come out of Mozilla, and as with anything one cares about, we are frustrated about some things. Such is life.
But with every new test harness being talked about and with every new set of tests being brought in, the same questions and issues seem to come up for me. Every new test harness I see being developed for MoCo technologies bring me hope, but it is rarely clear why the last one is not being talked about. The same issues seem to be stumbled over. What happened to the lessons learned in the past? One problem with wiki-based communities is that people talk a lot about what they are doing. People tend not to talk about what they are not doing. Search for any kind of solution in Mozilla's information space, in any problem area, and one finds at least four or five different efforts. Only one or two may be current, but nothing gets written about why the others are not moving forward, or why they failed. It may just be against our natures to talk about things that did not work. One just has to puzzle out the things that are real from the things that just look real.
In testing, it often seems to me that one needs to be clear about what kind of tests one wants, who can create, develop, or maintain what kinds of tests, and why different kinds of tests should or should not be run. If one is not clear about these things, tests get developed which do not get run, tests get run but they do not exercise the right things, or their results get ignored, and lots of people might be working hard to no good effect. The result is that there is no joy.
Questions:
(How hard is it to/Who can) develop a new test?
(How hard is it to/Who can) modify an existing test?
(How hard is it to/Who can) identify a feature or fault for which a test should be created?
(How hard is it to/Who can) evaluate the benefit of creating a test of the effectiveness of running a particular test over time?
(How hard is it to/Who can) run a test?
(How hard is it to/Who can) see that a test has failed?
(How hard is it to/Who can) identify the cause of a test failure, or interpret the failure?
I would suggest that each of these questions defines an axis in the "problem space", the space of possibilities to in which to locate problems and suggest solutions about Mozilla products. There are risks to be managed in the Firefox source base, and possibilities to be enabled. The cost of the risk to be averted and the possible improvement to the product that are enabled have to be measured against the costs of testing. I think that answers to these questions can be used to categorize various quality efforts and give a way to judge whether testing is doing what it should do and how much should or should not be spent on it.
I was going to take a stab at answering the one or more of questions above, but I will post this as a start. Answers can come later, especially since I do not know all of the answers. Nor am I sure I know all the questions. And if anyone wants to suggest either questions or answers, please do. And if anyone can think of a way to draw a representation of a, say, 12-dimensional space, so I can draw pictures based on something like the questions above, let me know that too.
Thanks. More to come.
But with every new test harness being talked about and with every new set of tests being brought in, the same questions and issues seem to come up for me. Every new test harness I see being developed for MoCo technologies bring me hope, but it is rarely clear why the last one is not being talked about. The same issues seem to be stumbled over. What happened to the lessons learned in the past? One problem with wiki-based communities is that people talk a lot about what they are doing. People tend not to talk about what they are not doing. Search for any kind of solution in Mozilla's information space, in any problem area, and one finds at least four or five different efforts. Only one or two may be current, but nothing gets written about why the others are not moving forward, or why they failed. It may just be against our natures to talk about things that did not work. One just has to puzzle out the things that are real from the things that just look real.
In testing, it often seems to me that one needs to be clear about what kind of tests one wants, who can create, develop, or maintain what kinds of tests, and why different kinds of tests should or should not be run. If one is not clear about these things, tests get developed which do not get run, tests get run but they do not exercise the right things, or their results get ignored, and lots of people might be working hard to no good effect. The result is that there is no joy.
Questions:
(How hard is it to/Who can) develop a new test?
(How hard is it to/Who can) modify an existing test?
(How hard is it to/Who can) identify a feature or fault for which a test should be created?
(How hard is it to/Who can) evaluate the benefit of creating a test of the effectiveness of running a particular test over time?
(How hard is it to/Who can) run a test?
(How hard is it to/Who can) see that a test has failed?
(How hard is it to/Who can) identify the cause of a test failure, or interpret the failure?
I would suggest that each of these questions defines an axis in the "problem space", the space of possibilities to in which to locate problems and suggest solutions about Mozilla products. There are risks to be managed in the Firefox source base, and possibilities to be enabled. The cost of the risk to be averted and the possible improvement to the product that are enabled have to be measured against the costs of testing. I think that answers to these questions can be used to categorize various quality efforts and give a way to judge whether testing is doing what it should do and how much should or should not be spent on it.
I was going to take a stab at answering the one or more of questions above, but I will post this as a start. Answers can come later, especially since I do not know all of the answers. Nor am I sure I know all the questions. And if anyone wants to suggest either questions or answers, please do. And if anyone can think of a way to draw a representation of a, say, 12-dimensional space, so I can draw pictures based on something like the questions above, let me know that too.
Thanks. More to come.
Wednesday, October 10, 2007
Extension Development, and Trials and Tribulations Thereof
Well, I am almost done with an extension I am developing for a third-party company. Wow! Was this ever harder than I thought it would be....
I had some slightly complicated things in the extension. 2.5K lines of javascript, a custom XPCOM component, a status bar icon, a toolbar popup button, a sidebar, and a partridge in a pear tree thrown in for good measure. The app was crashing on exit. Actually only FF 2.* was crashing. FF 3.* was fine.
So, I am doing a bunch of different kinds of allocations and there are listeners and so on. I figured I was doing something wrong with the memory. Much staring at the code. Much re-reading of dbaron's docs on this stuff. Much more staring at code....
Turned out it was this:
I thought that all this memory stuff might have a problem, and it turns out to be something bone simple! It is amazing what taking a machete to your code can show you. Along the way, an amusing thing happened. I took the image attribute off the toolbarbutton and I got this:
And someone said this was ok! I might try to guess at the logic, but I cannot make myself do it. Obviously, some people are very immersed in how XUL gets rendered. I hope I can learn enough to understand it. I hope I never forget that its contortions are, after all, contortions.
Next things to do now that this monkey is almost off my back:
- do an extension for rule-based qualification of XUL, because you need something to tell you that you can have A, B, and C together, but C makes B have no effect, and C without A does something totally different. Look at a given XUL reference. It inherits all its parent's attributes. Now, how many have any effect? How many break everything? There has got to be a way to get information about this.
- figure out something to help me search bugzilla. I looked for bugs on this many times and did not find it. Maybe it is because the relevant bug was marked for Windows. Maybe it was because the title mentioned <menupopup> and not <toolbarbutton>. Maybe it was because the title said it happened on window close and I was searching for "quit". I do not know. I just know I truly hate trying to find things in bugzilla. But when I suggest changes to bugzilla, they seem to hate that even more. I have no idea why. So, I just obviously need to construct tools to help me get around this thing. There has got to be a better way to find bugs than creating new bugs and letting gavin find the dupes. For one thing, this is not fair to gavin, or neil, or reed, or whoever it is.
- and some other things I am sure I have forgotten. We'll see.
I had some slightly complicated things in the extension. 2.5K lines of javascript, a custom XPCOM component, a status bar icon, a toolbar popup button, a sidebar, and a partridge in a pear tree thrown in for good measure. The app was crashing on exit. Actually only FF 2.* was crashing. FF 3.* was fine.
So, I am doing a bunch of different kinds of allocations and there are listeners and so on. I figured I was doing something wrong with the memory. Much staring at the code. Much re-reading of dbaron's docs on this stuff. Much more staring at code....
Turned out it was this:
<toolbarbutton id="id1" type="menu" popup="id2">
<menupopup id="id2" />
</toolbarbutton>
I thought that all this memory stuff might have a problem, and it turns out to be something bone simple! It is amazing what taking a machete to your code can show you. Along the way, an amusing thing happened. I took the image attribute off the toolbarbutton and I got this:
And someone said this was ok! I might try to guess at the logic, but I cannot make myself do it. Obviously, some people are very immersed in how XUL gets rendered. I hope I can learn enough to understand it. I hope I never forget that its contortions are, after all, contortions.
Next things to do now that this monkey is almost off my back:
- do an extension for rule-based qualification of XUL, because you need something to tell you that you can have A, B, and C together, but C makes B have no effect, and C without A does something totally different. Look at a given XUL reference. It inherits all its parent's attributes. Now, how many have any effect? How many break everything? There has got to be a way to get information about this.
- figure out something to help me search bugzilla. I looked for bugs on this many times and did not find it. Maybe it is because the relevant bug was marked for Windows. Maybe it was because the title mentioned <menupopup> and not <toolbarbutton>. Maybe it was because the title said it happened on window close and I was searching for "quit". I do not know. I just know I truly hate trying to find things in bugzilla. But when I suggest changes to bugzilla, they seem to hate that even more. I have no idea why. So, I just obviously need to construct tools to help me get around this thing. There has got to be a better way to find bugs than creating new bugs and letting gavin find the dupes. For one thing, this is not fair to gavin, or neil, or reed, or whoever it is.
- and some other things I am sure I have forgotten. We'll see.
Sunday, September 30, 2007
Bad Intel-Only Icon for Minefield
Someone posted a while back about this icon. I thought they said they were not sure what caused it. I got one of these and saw what it was.
This is what happens when you have an app that is built only for Intel, and you are looking at it on a PPC system. It is not going to run. If it is built "fat", then it has both an Intel executable and a PPC executable hidden inside its bundle directory and it will run on either.
This particular app was the Mac OS X 10.4.x app put up by Chris Double for his SVG Video Demo.
Just FYI.
This is what happens when you have an app that is built only for Intel, and you are looking at it on a PPC system. It is not going to run. If it is built "fat", then it has both an Intel executable and a PPC executable hidden inside its bundle directory and it will run on either.
This particular app was the Mac OS X 10.4.x app put up by Chris Double for his SVG Video Demo.
Just FYI.
Monday, September 10, 2007
Discovering the Coolness of nsIWebProgressListener
I am writing an extension and I have been jumping through some painful hoops with the "load" and "unload" events and "onDocumentLoad" and I have been trying to figure out how this is supposed to behave....
Then I discovered nsIWebProgressListener. What a beautiful interface!
I had thought that, in all my questions and grumblings and gripes, that someone would have pointed me to this. Someone probably did and I probably did not hear. But then, I notice that I will be drowning in a pool and I will say "Help!" and someone will say "you probably need an instance of a nsINylonLinearExtractionDevice...". No, actually, I need to not drown! Or I need a rope. And if I decode the answer, I can see that, yes, I was told I should grab a rope. How could I have missed it? :-)
It also makes it challenging that when you do a search (in site:mozilla.org), you are as likely to find a document from 1999 as any other. Yes, people have written new documents on the new ways of doing things. How often do the old documents get removed from wiki.m.o, or www.m.o? From the evidence, I would guess it does not happen very often.
There needs to be many more snippets on MDC that make use of this interface. I look forward to posting some. I am recalling many newbie questions on the newsgroups where the appropriate response would have been that one needs to look at all the wonderful things this interface gives you.
Of course, I always have a suggestion for future improvement. It's just something I have to do.
I wanted to keep track of some resources, one for each open window or tab. The interface, through the onStateChange and onLocationChange, gives me almost everything I need. Except that when I close a window, or close a tab, I get no message? I could just keep my stuff around in case a notification ever comes in on that window. But if that window was closed, wouldn't that be a leak in my extension?
I am sure there are other, older events I could look for, but it seems that nsIWebProgressListener was written to provide a better interface. It just needs to do the whole job.
See: http://bugzilla.mozilla.org/show_bug.cgi?id=395709 -
nsIWebProgressListener can send message when window or tab is closed
Then I discovered nsIWebProgressListener. What a beautiful interface!
I had thought that, in all my questions and grumblings and gripes, that someone would have pointed me to this. Someone probably did and I probably did not hear. But then, I notice that I will be drowning in a pool and I will say "Help!" and someone will say "you probably need an instance of a nsINylonLinearExtractionDevice...". No, actually, I need to not drown! Or I need a rope. And if I decode the answer, I can see that, yes, I was told I should grab a rope. How could I have missed it? :-)
It also makes it challenging that when you do a search (in site:mozilla.org), you are as likely to find a document from 1999 as any other. Yes, people have written new documents on the new ways of doing things. How often do the old documents get removed from wiki.m.o, or www.m.o? From the evidence, I would guess it does not happen very often.
There needs to be many more snippets on MDC that make use of this interface. I look forward to posting some. I am recalling many newbie questions on the newsgroups where the appropriate response would have been that one needs to look at all the wonderful things this interface gives you.
Of course, I always have a suggestion for future improvement. It's just something I have to do.
I wanted to keep track of some resources, one for each open window or tab. The interface, through the onStateChange and onLocationChange, gives me almost everything I need. Except that when I close a window, or close a tab, I get no message? I could just keep my stuff around in case a notification ever comes in on that window. But if that window was closed, wouldn't that be a leak in my extension?
I am sure there are other, older events I could look for, but it seems that nsIWebProgressListener was written to provide a better interface. It just needs to do the whole job.
See: http://bugzilla.mozilla.org/show_bug.cgi?id=395709 -
nsIWebProgressListener can send message when window or tab is closed
Saturday, September 01, 2007
Tools for UI Testing Coming Around...
In case anyone missed it, as I almost did, a technology that had been called Koala was released by IBM as CoScripter. Here is the post. I, for one, am jazzed about this.
There is an opportunity here that I want to try to take advantage of. It has been possible, for at least a few months, to use JavaScript to create events and throw them at a DOM tree and see what happens. Some really good work went into making this possible.
But it has been made possible. It has not been made easy. As far as I know, there has not been a flood of new mochitest test cases using this API to test the UI in Firefox. Which is a shame. There is a lot of good testing that needs to be done.
Right now, most UI testing is done via the "clicks and eyeballs" harness that is housed at Building K of MoCo Central. These people do a great job with what they are given, but frankly they should not have to do it all. If the UI was more specified and the development of automated tests was part of feature development, there would be less fire drills at release times. Fire drills may be good for building esprit de corps in an organization, but it does not always lead to a quality release. When we hear of a successful fire drill at MoCo, we should be glad they got through it, but we should be worried that it happened.
What does this have to do with CoScripter? Am I done with my MoCo org rant? Yes, I am. And let me tell you.
Last time I heard, CoScripter records JavaScript events. I still have to confirm this. A lot may have happened since this was shown at MoCo in February and it was said that it would be released "any day now" and then "when IBM legal signed off". And hey, it only took six months. Those lawyers must have really burned the midnight oil, yes? :-)
The point is that it may be possible to record a user doing stuff, and then turn that into an automated test case. I am sure some processing of the resulting recording will have to occur, but I have done automated test case generation with a UI recording tool at Apple (that is proprietary and shall remain nameless), so I am pretty sure it can be done.
Imagine a user being able to turn on an extension, and then do something that causes whatever error they are seeing, and then just drop that into a bug. It would certainly make user's bugs a bit more informative. And more automated UI tests? This cannot be a bad thing.
Hopefully I can do this and not fail my classes this semester. We will see. :-)
There is an opportunity here that I want to try to take advantage of. It has been possible, for at least a few months, to use JavaScript to create events and throw them at a DOM tree and see what happens. Some really good work went into making this possible.
But it has been made possible. It has not been made easy. As far as I know, there has not been a flood of new mochitest test cases using this API to test the UI in Firefox. Which is a shame. There is a lot of good testing that needs to be done.
Right now, most UI testing is done via the "clicks and eyeballs" harness that is housed at Building K of MoCo Central. These people do a great job with what they are given, but frankly they should not have to do it all. If the UI was more specified and the development of automated tests was part of feature development, there would be less fire drills at release times. Fire drills may be good for building esprit de corps in an organization, but it does not always lead to a quality release. When we hear of a successful fire drill at MoCo, we should be glad they got through it, but we should be worried that it happened.
What does this have to do with CoScripter? Am I done with my MoCo org rant? Yes, I am. And let me tell you.
Last time I heard, CoScripter records JavaScript events. I still have to confirm this. A lot may have happened since this was shown at MoCo in February and it was said that it would be released "any day now" and then "when IBM legal signed off". And hey, it only took six months. Those lawyers must have really burned the midnight oil, yes? :-)
The point is that it may be possible to record a user doing stuff, and then turn that into an automated test case. I am sure some processing of the resulting recording will have to occur, but I have done automated test case generation with a UI recording tool at Apple (that is proprietary and shall remain nameless), so I am pretty sure it can be done.
Imagine a user being able to turn on an extension, and then do something that causes whatever error they are seeing, and then just drop that into a bug. It would certainly make user's bugs a bit more informative. And more automated UI tests? This cannot be a bad thing.
Hopefully I can do this and not fail my classes this semester. We will see. :-)
Tuesday, July 31, 2007
What Prevents Multi-App Development at MoCo?
Reading some (but not all) of the Firefox/Thunderbird discussions, I find myself sanguine. I am not finding anything to be upset about. I think the current discussion may be more about the social structure of Mozilla. It does not seem to be a technical issue.
Looking at the technologies, it might be helpful if MoCo became, or came to see itself as, a service-provider for projects like Thunderbird. It could be "sourceforge with benefits", or perhaps a "sourceforge with a conscience". This seems obvious.
What makes it hard for MoCo to deal with Thunderbird as it is right now? Open-source avoids some problems one sees with commercial development. Commercial development seeks local maxima of utility with minimal investment of money. But MoCo has a similar issue. It seeks local maxima with minimal investment of social capital. What does this mean? If one looks at how Mozilla works, one sees this played out in many ways. I see some of them, but I do not think I am the best one to call them out.
One can identify, though, the "Mozilla way" of doing things. I noticed this when I was at Apple, and I still see it today. The "Mozilla way" is to use quirky tool sets, often old tools sets, with lots of hand-crafted modifications for particular issues. Enabling flexibility in these systems is not often a priority. Often it seems to be easier to hand-craft a solution to deal with the hand-crafted solution from two years ago. Finding a general solution that does not require tweaking does not seem to be a priority. A general solution rarely has social benefit. Individual tweaks, each to satisfy a different social entity, do have social benefits.
Why are more things not better documented? There is no social benefit to documenting that which is known by the core group. MoCo has, in general, a resistance to making things obvious. But social networks are built on shared knowledge, and if that knowledge requires a "maturation ritual", a "rite of passage", to find, all the better.
So, why can't Mozilla keep Thunderbird up-front, alongside Firefox? Perhaps it is not a technical issue. Anthropologists tell us that primates form social networks whose size varies with the ration of the brain size to body size. Given MoCo's reliance on social cohesion, Mozilla may not be able to concentrate on Thunderbird because the social network required is too big. Is Mozilla's pervasive reliance on social cohesion a good thing?
Looking at the technologies, it might be helpful if MoCo became, or came to see itself as, a service-provider for projects like Thunderbird. It could be "sourceforge with benefits", or perhaps a "sourceforge with a conscience". This seems obvious.
What makes it hard for MoCo to deal with Thunderbird as it is right now? Open-source avoids some problems one sees with commercial development. Commercial development seeks local maxima of utility with minimal investment of money. But MoCo has a similar issue. It seeks local maxima with minimal investment of social capital. What does this mean? If one looks at how Mozilla works, one sees this played out in many ways. I see some of them, but I do not think I am the best one to call them out.
One can identify, though, the "Mozilla way" of doing things. I noticed this when I was at Apple, and I still see it today. The "Mozilla way" is to use quirky tool sets, often old tools sets, with lots of hand-crafted modifications for particular issues. Enabling flexibility in these systems is not often a priority. Often it seems to be easier to hand-craft a solution to deal with the hand-crafted solution from two years ago. Finding a general solution that does not require tweaking does not seem to be a priority. A general solution rarely has social benefit. Individual tweaks, each to satisfy a different social entity, do have social benefits.
Why are more things not better documented? There is no social benefit to documenting that which is known by the core group. MoCo has, in general, a resistance to making things obvious. But social networks are built on shared knowledge, and if that knowledge requires a "maturation ritual", a "rite of passage", to find, all the better.
So, why can't Mozilla keep Thunderbird up-front, alongside Firefox? Perhaps it is not a technical issue. Anthropologists tell us that primates form social networks whose size varies with the ration of the brain size to body size. Given MoCo's reliance on social cohesion, Mozilla may not be able to concentrate on Thunderbird because the social network required is too big. Is Mozilla's pervasive reliance on social cohesion a good thing?
Tuesday, July 24, 2007
automated leak testing - followup
I have filed a bug (389361 - reproducible leak accessing page www.1ting.com) as a result of my leak testing. There are actually 6 URLs in that bug, but I cannot differentiate the leaks, so I have reported them together. I will have to figure out the leaks-gauge output so that I can make sense of this and report better bugs.
The leaks tool available on the Mac allows one to set an environment variable, MallocStackLogging, that makes the leaks output much more verbose. It will potentially tell you a lot about the objects being leaked. Unfortunately, this mostly works for Objective-C objects. C++ seems to be pretty good at obfuscating itself, or Apple has not done the work to help here, or both. It's a shame.
Well, the long and short of it is that leaks-gauge output does not make very much sense to me yet. :-(
I was not being very creative in how I got URLs. Someone pointed me to the alexa 500 and I hit pay dirt. Here are the sites I filed the bug against:
* http://www.1ting.com
* http://www.chinaren.com/
* http://www.cmfu.com/
* http://www.hurriyet.com.tr/
* http://www.yahoo.com/
* http://www.zaycev.net/
By the way, I am trying to get my add-on up onto AMO. This extension does nothing but quit Firefox at any page load. This is useful when launching Firefox from a script. The extension is available at my site and in the AMO sand box (login required, or may not be reachable). If anyone wants to review, after reading the docs, please do. I am not sure what the trigger is for getting it accepted on AMO. We will see.
I find it interesting that these leaks are sometimes not reproducible. There were a couple of dozen URLs in the list which generated leaks, but only once. As I re-tested URLs, sometimes they would leak every other time (like www.yahoo.com) and sometimes every third access (like www.1ting.com) and sometimes it was just more random than that. I did not list a URL if I could not get it to leak again. I am re-launching the browser each time, though, so I thought the testing would be more reproducible. It is more reproducible, but it is still surprising to me how many one-time leaks there were. There are some questions I am going to investigate.
* Does it matter if I randomize the list of URLs I am checking?
* If I automatically retry leaking URLs a certain number of times, how many times should I check?
* What files in the app bundle or in the profile directory change with an instance launch and shutdown? Or,
* What state being held between app launches?
We'll see.
The leaks tool available on the Mac allows one to set an environment variable, MallocStackLogging, that makes the leaks output much more verbose. It will potentially tell you a lot about the objects being leaked. Unfortunately, this mostly works for Objective-C objects. C++ seems to be pretty good at obfuscating itself, or Apple has not done the work to help here, or both. It's a shame.
Well, the long and short of it is that leaks-gauge output does not make very much sense to me yet. :-(
I was not being very creative in how I got URLs. Someone pointed me to the alexa 500 and I hit pay dirt. Here are the sites I filed the bug against:
* http://www.1ting.com
* http://www.chinaren.com/
* http://www.cmfu.com/
* http://www.hurriyet.com.tr/
* http://www.yahoo.com/
* http://www.zaycev.net/
By the way, I am trying to get my add-on up onto AMO. This extension does nothing but quit Firefox at any page load. This is useful when launching Firefox from a script. The extension is available at my site and in the AMO sand box (login required, or may not be reachable). If anyone wants to review, after reading the docs, please do. I am not sure what the trigger is for getting it accepted on AMO. We will see.
I find it interesting that these leaks are sometimes not reproducible. There were a couple of dozen URLs in the list which generated leaks, but only once. As I re-tested URLs, sometimes they would leak every other time (like www.yahoo.com) and sometimes every third access (like www.1ting.com) and sometimes it was just more random than that. I did not list a URL if I could not get it to leak again. I am re-launching the browser each time, though, so I thought the testing would be more reproducible. It is more reproducible, but it is still surprising to me how many one-time leaks there were. There are some questions I am going to investigate.
* Does it matter if I randomize the list of URLs I am checking?
* If I automatically retry leaking URLs a certain number of times, how many times should I check?
* What files in the app bundle or in the profile directory change with an instance launch and shutdown? Or,
* What state being held between app launches?
We'll see.
Thursday, July 19, 2007
automating leaks testing
I have managed to get some ideas I had to work in code. It required a couple of things.
First, why. The leak-gauge.pl script by dbaron is obviously very useful for tracking leaks in Firefox. I was just reminded of this reading this blog post by Steve England.
But reading Steve's post, it looks as though he did the following:
1) invoke leaks-gauge.pl on some URL
2) do a bunch of stuff on the resulting page
3) quit the browser
4) look at the results
5) perhaps file a bug
6) lather, rinse and report
If one wants to automate this testing, there are issues from some of this.
"invoke leaks-gauge.pl on some URL"
Right now, one uses a web page to tell leak-gauge.pl what to do. It is a manual process, hence this is not automatable.. I modified the leaks-gauge.pl script so that you give it an application executable, a profile directory, and a URL and it launches the browser for you. Another script goes through a list of apps and a list of URLs and calls the modified leaks-gauge.pl with the Cartesian product of these. This could all be done smarter than I am doing it.
What URLs can I run it on? Not wanting to answer that right now, I just looked at my copy of the trunk and used all the html files inside the layout/reftests directory. Right now this is 1081 URLs. I used 5 different applications, those being nightlies from 2007/01/30, 2007/03/24, 2007/05/23, 2007/06/15, and 2007/07/13.
"do a bunch of stuff"
Steve was testing extensions. Indeed, if one wants to test the memory footprint of extensions, one probably has to interact with the UI to cause the extension to do its stuff. I have no solutions to this quandary. I can say that it would be useful to use leak-gauge.pl just for single-page loads and this is automatable now.
"quit the browser"
This is not hard, but it is a bother. There are lots of ways to do it, but many of them make testing hard for various reasons. My solution to this was to create a very small extension, which will be up on AMO very soon. All it does is wait for a page to load, any page, and quit the browser. So, if one invokes firefox from the command-line and passes in a URL, then Firefox gets launched, it loads the page and then it automatically quits.
"look at the results"
So, I have a script that takes the output from the test run and outputs a couple of pages. Of course, anybody would know I will say this should be in a database, since I think everything should be in a database, but that is for later.
For full results, see http://www.wykiwyk.com/mozilla/leakTesting/testlog_20070718_byApp.html and http://www.wykiwyk.com/mozilla/leakTesting/testlog_20070718_byURL.html.
The summary results are that 2007/01/30 leaked some on a lot of pages, 2007/03/24 was slightly worse, and 2007/05/23, 2007/06/15, and 2007/07/13 did not leak at all on these files. Yeah for Cycle Collection! :-)
And now, there is a Firefox extension for which someone is actually paying me, so I should get back to it.
First, why. The leak-gauge.pl script by dbaron is obviously very useful for tracking leaks in Firefox. I was just reminded of this reading this blog post by Steve England.
But reading Steve's post, it looks as though he did the following:
1) invoke leaks-gauge.pl on some URL
2) do a bunch of stuff on the resulting page
3) quit the browser
4) look at the results
5) perhaps file a bug
6) lather, rinse and report
If one wants to automate this testing, there are issues from some of this.
"invoke leaks-gauge.pl on some URL"
Right now, one uses a web page to tell leak-gauge.pl what to do. It is a manual process, hence this is not automatable.. I modified the leaks-gauge.pl script so that you give it an application executable, a profile directory, and a URL and it launches the browser for you. Another script goes through a list of apps and a list of URLs and calls the modified leaks-gauge.pl with the Cartesian product of these. This could all be done smarter than I am doing it.
What URLs can I run it on? Not wanting to answer that right now, I just looked at my copy of the trunk and used all the html files inside the layout/reftests directory. Right now this is 1081 URLs. I used 5 different applications, those being nightlies from 2007/01/30, 2007/03/24, 2007/05/23, 2007/06/15, and 2007/07/13.
"do a bunch of stuff"
Steve was testing extensions. Indeed, if one wants to test the memory footprint of extensions, one probably has to interact with the UI to cause the extension to do its stuff. I have no solutions to this quandary. I can say that it would be useful to use leak-gauge.pl just for single-page loads and this is automatable now.
"quit the browser"
This is not hard, but it is a bother. There are lots of ways to do it, but many of them make testing hard for various reasons. My solution to this was to create a very small extension, which will be up on AMO very soon. All it does is wait for a page to load, any page, and quit the browser. So, if one invokes firefox from the command-line and passes in a URL, then Firefox gets launched, it loads the page and then it automatically quits.
"look at the results"
So, I have a script that takes the output from the test run and outputs a couple of pages. Of course, anybody would know I will say this should be in a database, since I think everything should be in a database, but that is for later.
For full results, see http://www.wykiwyk.com/mozilla/leakTesting/testlog_20070718_byApp.html and http://www.wykiwyk.com/mozilla/leakTesting/testlog_20070718_byURL.html.
The summary results are that 2007/01/30 leaked some on a lot of pages, 2007/03/24 was slightly worse, and 2007/05/23, 2007/06/15, and 2007/07/13 did not leak at all on these files. Yeah for Cycle Collection! :-)
And now, there is a Firefox extension for which someone is actually paying me, so I should get back to it.
Saturday, July 07, 2007
Starting a Firefox Instance With a Fresh Profile
Maybe it is just me, but in doing anything with development builds of Firefox, I was constantly bothered by having my profile mucked around with, or having to create fresh profiles, or having to switch the default profile for some set of tests I wanted to run.
I had slapped together little scripts to deal with some of this, but I finally got them all together in one script and it has been working for me for a while. This may be useful to others. Maybe not.
One thing different about this is that I am doing things on a Mac. I have not tried to make the script cross-platform at all. I also find the executables in the places I keep them. There is no way to find out about the different copies of Firefox that are installed on a particular machine. There is no global installation log. And even though all of the apps read from resources in the 'Application Support' folder, none of them store information there that would tell one which instances of Firefox have been run. And it is not the most elegnt or clever script I have written. But it works. If anybody has suggestions for something to replace this find command, or anything else, please suggest. Until then, here it is:
http://www.wykiwyk.com/mozilla/newbrowser.pl
So, the script will create a fresh profile, put some preferences into it, ask you to identify the copy of Firefox that you want to run and then launch with the new profile. Here is an example of it being run:
I have also made it trivial to run a set of reftests, as so:
For more:
I had slapped together little scripts to deal with some of this, but I finally got them all together in one script and it has been working for me for a while. This may be useful to others. Maybe not.
One thing different about this is that I am doing things on a Mac. I have not tried to make the script cross-platform at all. I also find the executables in the places I keep them. There is no way to find out about the different copies of Firefox that are installed on a particular machine. There is no global installation log. And even though all of the apps read from resources in the 'Application Support' folder, none of them store information there that would tell one which instances of Firefox have been run. And it is not the most elegnt or clever script I have written. But it works. If anybody has suggestions for something to replace this find command, or anything else, please suggest. Until then, here it is:
http://www.wykiwyk.com/mozilla/newbrowser.pl
So, the script will create a fresh profile, put some preferences into it, ask you to identify the copy of Firefox that you want to run and then launch with the new profile. Here is an example of it being run:
% newbrowser
1: /Users/ray/mo/trowser/mozilla/dist/MinefieldDebug.app/Contents/MacOS/firefox
2: /Users/ray/mo/trowser2/mozilla/dist/Minefield.app/Contents/MacOS/firefox
3: /Applications/Mozilla/Firefox_1.5.0.11.app/Contents/MacOS/firefox
4: /Applications/Mozilla/Firefox_1.5.app/Contents/MacOS/firefox
5: /Applications/Mozilla/Firefox_2.0.0.2.app/Contents/MacOS/firefox
6: /Applications/Mozilla/Firefox_2.0.0.4.app/Contents/MacOS/firefox
7: /Applications/Mozilla/GranParadiso_20070426.app/Contents/MacOS/firefox
8: /Applications/Mozilla/Minefield_3.0a2pre_20070130.app/Contents/MacOS/firefox
9: /Applications/Mozilla/Minefield_3.0a2pre_20070206.app/Contents/MacOS/firefox
10: /Applications/Mozilla/Minefield_3.0a3pre_20070222.app/Contents/MacOS/firefox
11: /Applications/Mozilla/Minefield_3.0a3pre_20070308.app/Contents/MacOS/firefox
12: /Applications/Mozilla/Minefield_3.0a3pre_20070317.app/Contents/MacOS/firefox
13: /Applications/Mozilla/Minefield_3.0a3pre_20070324.app/Contents/MacOS/firefox
14: /Applications/Mozilla/Minefield_3.0a4pre_20070413.app/Contents/MacOS/firefox
15: /Applications/Mozilla/Minefield_3.0a5pre_20070503.app/Contents/MacOS/firefox
16: /Applications/Mozilla/Minefield_3.0a5pre_20070519.app/Contents/MacOS/firefox
17: /Applications/Mozilla/Minefield_3.0a6pre_20070623.app/Contents/MacOS/firefox
which executable? 1
cmd = "NO_EM_RESTART=1 /Users/ray/mo/trowser_newbad/mozilla/dist/Minefield.app/Contents/MacOS/firefox -profile /tmp/3lwhsxqz.mozilla_20070707_215451_PDT 2>&1"
WARNING: NS_ENSURE_TRUE(compMgr) failed: file nsComponentManagerUtils.cpp, line 90
Type Manifest File: /tmp/3lwhsxqz.mozilla_20070707_215451_PDT/xpti.dat
*** Registering xpconnect components (all right -- a generic module!)
etc, etc, etc, etc, ....
I have also made it trivial to run a set of reftests, as so:
% newbrowser -reftest layout/reftests/reftest.list
1: /Users/ray/mo/trowser/mozilla/dist/MinefieldDebug.app/Contents/MacOS/firefox
2: /Users/ray/mo/trowser_newbad/mozilla/dist/Minefield.app/Contents/MacOS/firefox
3: /Applications/Mozilla/Firefox_1.5.0.11.app/Contents/MacOS/firefox
4: /Applications/Mozilla/Firefox_1.5.app/Contents/MacOS/firefox
5: /Applications/Mozilla/Firefox_2.0.0.2.app/Contents/MacOS/firefox
6: /Applications/Mozilla/Firefox_2.0.0.4.app/Contents/MacOS/firefox
7: /Applications/Mozilla/GranParadiso_20070426.app/Contents/MacOS/firefox
8: /Applications/Mozilla/Minefield_3.0a2pre_20070130.app/Contents/MacOS/firefox
9: /Applications/Mozilla/Minefield_3.0a2pre_20070206.app/Contents/MacOS/firefox
10: /Applications/Mozilla/Minefield_3.0a3pre_20070222.app/Contents/MacOS/firefox
11: /Applications/Mozilla/Minefield_3.0a3pre_20070308.app/Contents/MacOS/firefox
12: /Applications/Mozilla/Minefield_3.0a3pre_20070317.app/Contents/MacOS/firefox
13: /Applications/Mozilla/Minefield_3.0a3pre_20070324.app/Contents/MacOS/firefox
14: /Applications/Mozilla/Minefield_3.0a4pre_20070413.app/Contents/MacOS/firefox
15: /Applications/Mozilla/Minefield_3.0a5pre_20070503.app/Contents/MacOS/firefox
16: /Applications/Mozilla/Minefield_3.0a5pre_20070519.app/Contents/MacOS/firefox
17: /Applications/Mozilla/Minefield_3.0a6pre_20070623.app/Contents/MacOS/firefox
which executable? 1
cmd = "NO_EM_RESTART=1 /Users/ray/mo/trowser/mozilla/dist/MinefieldDebug.app/Contents/MacOS/firefox -profile /tmp/60yqqqnd.mozilla_20070707_221500_PDT --reftest layout/reftests/reftest.list 2>&1 | /usr/bin/grep '^REFTEST '"
etc, etc, etc, etc, ....
For more:
% newbrowser -usage
newbrowser [ -noUserJS ] [ -reftest <manifest file> ] [ -exec <executable> ]
%
Sunday, July 01, 2007
A Way Around my Gordian Knot with Reftests
I have been trying to figure out a way for reftests to be able to do image comparisons. I have been hampered by both my lack of experience with writing "Mozilla-ish" JavaScript and by the fact that reftest was not designed to do this. Or perhaps, reftest was designed not to do this. Indeed, the thing that makes reftest different is that it does not do exact comparisons with a hard-coded "golden" page and so it is more flexible and does not cause unnecessary failures when unimportant changes are made to, for example, layout.
But there is a call for exact rendering comparisons.
For example, if a certain Unicode value in an html page is supposed to render an Urdu character, the only way to test for this is to have a human look at it and see whether it is right or not. Similarly, whether a given MathML leads to correctly rendered expressions is, at this point, only checkable by a human loading a page and looking at it. I call this the "browsers and eyeballs" test harness.
But the people blocking my changes for reftest are better at doing that than I am at figuring out what would be acceptable and would also do the job, so I have figured out a way to use the current, checked-in version of reftest to do exact visual comparisons. Rather than go into the code, I will demonstrate.
First, I created a file, u1.txt, that contained a list of URLs that I pulled from Ian Hickson's "HTML 4.0 Test Suite" at http://hixie.ch/tests/html40/test41-1a.html . Following the chain that starts at that page leads one to 16 URLs.
Then I run:
The profile is nothing special. It is just dynamically generated so that the browser can run in an automation-friendly manner. Then u2.txt will contain something like:
Then when I can do:
This will verify that the html being received from the URL is exactly as is expected (the DGT marker above is for an SHA1 digest of the source) and the new copy of Firefox displays exactly the same images as the old copy.
Are there obvious issues raised by doing this kind of testing? Yes. And there are ways around them. There are some tools needed to support a workflow that uses this test tool. And I could try to guess what the objections will be now but, then again, "vita brevis, programmi longa".
This should definitely not be seen as an attempt to replace testing tools like mochitest or reftest. On the other hand, I could spend a few days and find 50,000 static URLs and be able to say whether any given check-in causes any difference in the display of any of those pages. How long would it take to find the 5000 or so important differences in those pages and write reftests for them? I do not know, but I am not going to spend a few years trying to find out.
But there is a call for exact rendering comparisons.
For example, if a certain Unicode value in an html page is supposed to render an Urdu character, the only way to test for this is to have a human look at it and see whether it is right or not. Similarly, whether a given MathML leads to correctly rendered expressions is, at this point, only checkable by a human loading a page and looking at it. I call this the "browsers and eyeballs" test harness.
But the people blocking my changes for reftest are better at doing that than I am at figuring out what would be acceptable and would also do the job, so I have figured out a way to use the current, checked-in version of reftest to do exact visual comparisons. Rather than go into the code, I will demonstrate.
First, I created a file, u1.txt, that contained a list of URLs that I pulled from Ian Hickson's "HTML 4.0 Test Suite" at http://hixie.ch/tests/html40/test41-1a.html . Following the chain that starts at that page leads one to 16 URLs.
Then I run:
% perl runRef -generate -src u1.txt -app /Users/ray/mozilla/dist/MinefieldDebug.app/Contents/MacOS/firefox -profile /tmp/12qp5rt8.mozilla_20070630_153321_PDT > u2.txt
The profile is nothing special. It is just dynamically generated so that the browser can run in an automation-friendly manner. Then u2.txt will contain something like:
URL: http://hixie.ch/tests/html40/test41-1a.html
DGT: 9e747d8a7b957f5126f07135b75f9564405da513
URI: data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAyAAAAPoCAYAAAAmy5qxAAAgAElEQVR4nOzdf1zV9 (and so on ...)
URL: http://hixie.ch/tests/html40/test41-1b.html
DGT: 1bb9c5fba9fd37d6afdced89817f0a61104b3ced
URI: data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAyAAAAPoCAYAAAAmy5qxAAAgAElEQVR4nOzdf1zV9d3/8QfXgA4 (and so on ...)
URL: http://hixie.ch/tests/html40/test41-2.shtml
DGT: 0f3e0c63e174afff4c61924384cbe2dc78722447
URI: data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAyAAAAPoCAYAAAAmy5qxAAAgAElEQVR4nOzdf1zV9d3/8QfXAA8aB (and so on ...)
URL: http://hixie.ch/tests/html40/test41-3.html
DGT: ad782efed8ab827701fb6740f2089d897f8499bd
URI: data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAyAAAAPoCAYAAAAmy5qxAAAgAElEQVR4nOzdf1zV9f (and so on ...)
Then when I can do:
% <re-fetch and build Firefox here>
% perl runRef -test -src u2.txt -app /Users/ray/mozilla/dist/MinefieldDebug.app/Contents/MacOS/firefox -profile /tmp/12qp5rt8.mozilla_20070630_153321_PDT
URL: http://hixie.ch/tests/html40/test41-1a.html -> Ok
URL: http://hixie.ch/tests/html40/test41-1b.html -> Ok
URL: http://hixie.ch/tests/html40/test41-2.shtml -> Ok
URL: http://hixie.ch/tests/html40/test41-3.html -> Ok
...
This will verify that the html being received from the URL is exactly as is expected (the DGT marker above is for an SHA1 digest of the source) and the new copy of Firefox displays exactly the same images as the old copy.
Are there obvious issues raised by doing this kind of testing? Yes. And there are ways around them. There are some tools needed to support a workflow that uses this test tool. And I could try to guess what the objections will be now but, then again, "vita brevis, programmi longa".
This should definitely not be seen as an attempt to replace testing tools like mochitest or reftest. On the other hand, I could spend a few days and find 50,000 static URLs and be able to say whether any given check-in causes any difference in the display of any of those pages. How long would it take to find the 5000 or so important differences in those pages and write reftests for them? I do not know, but I am not going to spend a few years trying to find out.
Thursday, April 19, 2007
Proposed Litmus Database Schema
I had proposed this change to the guys who run the litmus site. They had not responded, as they are in the middle of other changes. The other changes are, I think, orthogonal to these suggestions. I wanted to capture this suggestion before I lost it.
I think that this is the current structure (as I have reverse-engineered it) of the Litmus database.
I am proposing this, which I think will be much more flexible and will scale better in the long run. I am seeing a few slight errors, such as that there should be "0+" resources associated with a test, not "1+". But I think this is mostly correct.
The Category entity and its self-link allows a flexible categorization of the tests, but also allows a heirarchical categorization at the same time. The Resource entity would allow us to capture, for example, a URL as one resource of a "go to site of URL" test. Then associating 100 URLs with that test would be more flexible than creating 100 tests of the form "go to URL A", "go to URL B", etc. The Config entity would capture the type of machine and the versions of what is being run. This would be more flexible than, for example, creating a one version of test for Firefox 1.5, then a separate copy of the test for FireFox, then a separate copy of the test for Firefox 3, and so on.
If anyone has suggestions, please let me know.
I think that this is the current structure (as I have reverse-engineered it) of the Litmus database.
I am proposing this, which I think will be much more flexible and will scale better in the long run. I am seeing a few slight errors, such as that there should be "0+" resources associated with a test, not "1+". But I think this is mostly correct.
The Category entity and its self-link allows a flexible categorization of the tests, but also allows a heirarchical categorization at the same time. The Resource entity would allow us to capture, for example, a URL as one resource of a "go to site of URL" test. Then associating 100 URLs with that test would be more flexible than creating 100 tests of the form "go to URL A", "go to URL B", etc. The Config entity would capture the type of machine and the versions of what is being run. This would be more flexible than, for example, creating a one version of test for Firefox 1.5, then a separate copy of the test for FireFox, then a separate copy of the test for Firefox 3, and so on.
If anyone has suggestions, please let me know.
Thursday, April 12, 2007
Wednesday, April 04, 2007
A Random Attempt at Specifying Firefox UI
Correction: 2007/04/07 - This is why I like specs. I get things wrong at times. I was switching back and forth between FF and TB and wrote down the wrong menu headers. I am fixing the Firefox menu headers for Mac OS X. Thanks for the notes about this in the comments. This will be the last edit here. This document will be moved. Back to the original post....
I was telling people that this should be done, so I decided to put my feet into the dance. There is a lot of good stuff written in various blog entries and on wikis scattered hither and yon that describe various aspects of the UI of Gecko apps, such as Firefox. I mentioned that these could be brought together into some sort of specification that could be used to help people who are trying to test the product, document the product, develop for the product, or just plain understand it.
I borrowed a picture from Alex Faaborg, something I really wish I had seen in my first month at MoCo, when I was writing things in bugs that sounded like "that thing that does that stuff, next to that other thing, I do not think it is working...". Yeah. That went really well.
It is going to be difficult for one person to be comprehensive. I wanted to illustrate, though, an approach that can be taken. So, I have tried to describe the empty first page. I ended up also describing the "Quick Find" box. I think people know that command-F brings up Find, from the menu, but how many people know that the '/' key brings up the Quick Find bar instead? People remember vi, yes? Of course, they do. Very intuitive. :-)
So, here goes.
Firefox - A Browser
Of course, one does not always see these UI elements all at once. This will help more if it actually describes the state the browser is in at the time one sees this UI. While we're here, how did we get here? That is the big picture. But there are other pictures.
The first image above is the first page one sees on a Mac OS X system running Firefox 2.0.0.3 and the second is the same Firefox release on Ubuntu Linux 6.10. Obviously, there are differences between all the platforms.
- There is no Menu Bar in the window on Mac OS X. The menu bar for a Mac OS X application is at the top of the system window. The top-level menus are different.
* Mac top menus: File, Edit, View, History, Bookmarks, Tools, Window, Help
* Windows top menus: ???
* Mac top menus: File, Edit, View, History, Bookmarks, Tools, Help
- The Navigation Toolbar appears to be the same. Add-ons may place other icons in the Navigation Toolbar.
- In the Windows picture, there is a square box to the right of the Menu Bar. This is the "Busy Searching Spinner".
- The Mac has a "Busy Searching Spinner" on the right of the Search Bar. It spins when one is searching. It stays motionless while one is not searching.
- The Bookmarks Toolbar is the same.
- The Tabstrip is the same, except that the icons are different.
- The Content Area is the same. By default, the Content Area is blank. Firefix may be configured to open a page at launch. To see a blank Content Area, type "
about:blank" in the Location Bar.- The Sidebar is the same. It is only made visible under certain conditions. By default the browser history is made visible in the Sidebar when one types the 'History Sidebar Key'. Typing key again makes the History sidebar appear. Typing this key again makes it disappear.
* Mac - 'History Sidebar Key' = command-shift-h
* Windows - 'History Sidebar Key' = ???
* Linux - 'History Sidebar Key' = control-h
- The Status Bar is the same.
- The Find Bar is the same. Typing 'Find Key' makes the Find Bar visible below the Content Area and Sidebar.
* Mac: 'Find Key' = command-f
* Windows: 'Find Key' = ???
* Linux: 'Find Key' = control-f
- There is also a Quick Find Bar that replaces the Find Bar.
Quick Find Bar:
The Quick Find Bar in a browser window on Mac OS X is shown below.
If the Content Area is not blank, typing the '/' key makes the Quick Find Bar visible below the Content Area and Sidebar. There is also an issue with the default keyboard focus. For example, bring up the page shown below and type the '/' character. The Quick Find Bar will not appear because the default keyboard focus is on the search form field. One must click somewhere else on the page to take the keyboard focus away from the form field. When there is no form field accepting keyboard focus, then typing the '/' key brings up the Quick Find Bar.
If the Content Area is blank, typing the '/' key has no effect. The Quick Find Bar disappears after approximately 4 seconds if nothing is typed.
If one types a character, it is put into the 'Quick Find' field. If there is text on the page matching the text in the Quick Find text field, then the first instance of that text on the page is selected. If the selected text is part of a link, the URL for that link is displayed in the Status Bar and if one types a return key or an enter key, then that link is activated. If one types text which does not appear on the page, the background of the Quick Find text field turns red. If another character is typed and the action does not place a text string which is found on the page into the Quick Find text field, there is an audible "beep".
For example, if one navigates to the page above and types a '/' character and then a 'xxx', the result is two audible beeps and the page below.
If one does not hit return and stops typing, the Quick FInd Bar and the text typed in disappears. If one hits the '/' key again, the Quick Find Bar appears again, but the text field is empty and the previously typed text is lost.
Still To Be Done
Absolutely everything else.
Friday, March 02, 2007
Cranking on the Test Engine
I recently "finished" an inventory of the testing tools and test suites available here at Mozilla. I put quotes around the word "finished" because, of course, it is not done. In a sense, it never will be. What I have done is that really turned one crank on an engine. Hopefully, the engine will be started.
I view this as an engine because there is a cycle that needs to occur. (I may actually have my engine terminolgy wrong here. I do not actually know engines as anything other than an abstraction. Apologies in advance.) Also, I am not an artist. I do words, not pictures.
Spark - identify and ignite existing tests. This will lead to...
Combustion - tests are executed and results recorded.
There has been a lot of random firing in the past. There are a bunch of tests. But when people look to run tests, their choice of tests has been close to random. I believe this to be so because if one wanted to pick tests by another method, there is no way to do so. There is no information available to provide any other criteria. There is a very rough correspondence between modules and tests, but modules are amorphously defined and the results are usually passed on in the form of a message that "I ran tests in X and they worked." Not very specific.
Right now, combustion is starting to happen. Litmus tests get executed and there is information at the end that can be pointed to. There are suites of reftests and mochitests scripts that are being run after a build and reported in tinderbox. But there are a lot of tests that are not in these harnesses yet. There are a lot of tests that are hard to analyze and hard to understand. It is hard to tell how useful they are.
Expansion - results of testing are analyzed.
It is usually in this stage that a determination is made about whether something can be shipped, or if the quality of the product is deemed good enough for some purpose. This stage involves an evaluation about what the testing meant. This leads to the next stage.
Compression - functionality is identified for which new tests are needed.
Did the testing that was done test the features that we care about? Usually, enough testing is done so that we are confident of the new features and of drastically changed functionality. But we all know there are side-effects and side-effects to side-effects and I do not think anybody feels we are testing as much of the functionality as we should. Thinking about what new tests are needed pushes the piston back down to a firing position.
Injection - define more tests of the functionality.
And then this brings us back to Spark.
I suggest this could be thought of as a cycle, and a cycle that needs to continue to go around. It is not something to do until we are "done".
I view this as an engine because there is a cycle that needs to occur. (I may actually have my engine terminolgy wrong here. I do not actually know engines as anything other than an abstraction. Apologies in advance.) Also, I am not an artist. I do words, not pictures.
Spark - identify and ignite existing tests. This will lead to...
Combustion - tests are executed and results recorded.
There has been a lot of random firing in the past. There are a bunch of tests. But when people look to run tests, their choice of tests has been close to random. I believe this to be so because if one wanted to pick tests by another method, there is no way to do so. There is no information available to provide any other criteria. There is a very rough correspondence between modules and tests, but modules are amorphously defined and the results are usually passed on in the form of a message that "I ran tests in X and they worked." Not very specific.
Right now, combustion is starting to happen. Litmus tests get executed and there is information at the end that can be pointed to. There are suites of reftests and mochitests scripts that are being run after a build and reported in tinderbox. But there are a lot of tests that are not in these harnesses yet. There are a lot of tests that are hard to analyze and hard to understand. It is hard to tell how useful they are.
Expansion - results of testing are analyzed.
It is usually in this stage that a determination is made about whether something can be shipped, or if the quality of the product is deemed good enough for some purpose. This stage involves an evaluation about what the testing meant. This leads to the next stage.
Compression - functionality is identified for which new tests are needed.
Did the testing that was done test the features that we care about? Usually, enough testing is done so that we are confident of the new features and of drastically changed functionality. But we all know there are side-effects and side-effects to side-effects and I do not think anybody feels we are testing as much of the functionality as we should. Thinking about what new tests are needed pushes the piston back down to a firing position.
Injection - define more tests of the functionality.
And then this brings us back to Spark.
I suggest this could be thought of as a cycle, and a cycle that needs to continue to go around. It is not something to do until we are "done".
Subscribe to:
Posts (Atom)
