Community
Participate
Working Groups
(In reply to Arun Thondapu from bug 423561 comment #7) > In my local tests, the Browser uses Webkit but on > the actual test machine, it is using Mozilla Xulrunner, which may exhibit > different behavior. Since both XULRunner and WebKit are supported on GTK, the browser tests should be run on both browsers. This may require a separate target in test.xml that sets up the environment and then runs just the browser tests on the other browser.
Running browser tests in a separate VM (separate call to <ant target="core-test" ...>) would probably be a good idea anyway. To kill browser tests already after 10 minutes, add this to the Ant call: <property name="timeout" value="600000"/>
It would be nice having tests run with both webkit and xulrunner but testing the default would be better until someone spends time to improve the tests for running with both. And we would need the package providing libwebkitgtk-1.0.so.0 installed on the linux test boxes to even try enabling both so we have to request that. Adding webmaster on cc to arrange that.
I can see the following are installed on build.eclipse.org libwebkit-1.0.so.1 libwebkit-1.0.so.1.0.0 libwebkit-1.0.so.2 libwebkit-1.0.so.2.17.9 I'm not sure what would provide specificially "libwebkitgtk-1.0.so.0" but as far as I can tell build.eclipse.org has all the available libwebkit libraries installed except for the "-devel" package. S | Name | Summary | Type --+-----------------+----------------------------------------------+----------- | libwebkit | Library for rendering web content, GTK+ Port | srcpackage i | libwebkit-1_0-1 | Library for rendering web content, GTK+ Port | package i | libwebkit-1_0-2 | Library for rendering web content, GTK+ Port | package | libwebkit-devel | Library for rendering web content, GTK+ Port | package i | libwebkit-lang | Languages for package libwebkit | package | libwebkit1_0 | Library for rendering web content, GTK+ Port | srcpackage
(In reply to Thanh Ha from comment #3) > I can see the following are installed on build.eclipse.org > And ... what about the Hudson slaves? Are they the same? (or, use "build.eclipse.org" under the covers). In particular, I think we usually use 'slave 4' for these tests.
(In reply to Thanh Ha from comment #3) > I can see the following are installed on build.eclipse.org > > > libwebkit-1.0.so.1 > libwebkit-1.0.so.1.0.0 > libwebkit-1.0.so.2 > libwebkit-1.0.so.2.17.9 > Is libwebkit-1.0.so.2 smth that's newly installed? SWT will try to load this so file and this might explain why swt tests haven't failed lately - they run on webkit now?
(In reply to David Williams from comment #4) > (In reply to Thanh Ha from comment #3) > > I can see the following are installed on build.eclipse.org > > > > And ... what about the Hudson slaves? Are they the same? (or, use > "build.eclipse.org" under the covers). In particular, I think we usually use > 'slave 4' for these tests. Looks like webkit isn't installed at all on slave4. I went ahead and installed libwebkit-1_0-1 on slave4. Can we test before I go ahead and install this on all the slaves?
(In reply to Alexander Kurtakov from comment #5) > (In reply to Thanh Ha from comment #3) > > I can see the following are installed on build.eclipse.org > > > > > > libwebkit-1.0.so.1 > > libwebkit-1.0.so.1.0.0 > > libwebkit-1.0.so.2 > > libwebkit-1.0.so.2.17.9 > > > Is libwebkit-1.0.so.2 smth that's newly installed? SWT will try to load this > so file and this might explain why swt tests haven't failed lately - they > run on webkit now? No this came from the package "libwebkit-1_0-2" which I installed just before making that post. I wanted to see if it provided the so.0 file you were looking for. When I checked on slave4 now I noticed that package was not even available on slave4 so I might remove it from build.eclipse.org if to keep the versions the same.
(In reply to David Williams from comment #4) > And ... what about the Hudson slaves? Are they the same? (or, use > "build.eclipse.org" under the covers). In particular, I think we usually use > 'slave 4' for these tests. All the slaves are different hardware except for slave2 which is build.eclipse.org.
(In reply to Thanh Ha from comment #7) > (In reply to Alexander Kurtakov from comment #5) > > (In reply to Thanh Ha from comment #3) > > > I can see the following are installed on build.eclipse.org > > > > > > > > > libwebkit-1.0.so.1 > > > libwebkit-1.0.so.1.0.0 > > > libwebkit-1.0.so.2 > > > libwebkit-1.0.so.2.17.9 > > > > > Is libwebkit-1.0.so.2 smth that's newly installed? SWT will try to load this > > so file and this might explain why swt tests haven't failed lately - they > > run on webkit now? > > No this came from the package "libwebkit-1_0-2" which I installed just > before making that post. I wanted to see if it provided the so.0 file you > were looking for. > > When I checked on slave4 now I noticed that package was not even available > on slave4 so I might remove it from build.eclipse.org if to keep the > versions the same. Note that the one I looked for is named libwebkit*gtk* and it's the new name of the so file. Please don't remove whatever provides the libwebkit-1.0.so.2 as swt loads only: libwebkitgtk-3.0.so.0 libwebkit-1.0.so.2 libwebkitgtk-1.0.so.0
(In reply to Alexander Kurtakov from comment #9) > Note that the one I looked for is named libwebkit*gtk* and it's the new name > of the so file. Please don't remove whatever provides the libwebkit-1.0.so.2 > as swt loads only: > libwebkitgtk-3.0.so.0 > libwebkit-1.0.so.2 > libwebkitgtk-1.0.so.0 hmm ok in that case since the "libwebkit-1.0.so.2" didn't exist until about an hour ago then we will need to get that onto the hudson-slaves too so that they all have this file.
(In reply to Thanh Ha from comment #10) > hmm ok in that case since the "libwebkit-1.0.so.2" didn't exist until about > an hour ago then we will need to get that onto the hudson-slaves too so that > they all have this file. I spoke to Alexander on IRC and we decided to install this on all the slaves so slaves 1-6 now have libwebkit-1.0.so.2 installed. Let me know if there's anything else you need here.
(In reply to Thanh Ha from comment #11) > (In reply to Thanh Ha from comment #10) > > hmm ok in that case since the "libwebkit-1.0.so.2" didn't exist until about > > an hour ago then we will need to get that onto the hudson-slaves too so that > > they all have this file. > > I spoke to Alexander on IRC and we decided to install this on all the slaves > so slaves 1-6 now have libwebkit-1.0.so.2 installed. Let me know if there's > anything else you need here. Thanks Thanh! SWT: this sort of thing seems to be a recurring issue (such as from community, those leading the bleeding edge with Ubuntu 13.10, etc.) I wonder if we should have a clear statement of "what's required" (for both webkit and xulrunner varieties) and ideally some directions for "how to tell what's installed" since as far as I know, all the main distributions have different "techniques" and repositories, and library naming conventions. (dpkg vs. rpm, etc.). Anyway, I added "readme" to the keywords for that reason ... if you disagree or have a different method of documenting that, feel free to change it.
(In reply to Thanh Ha from comment #11) > (In reply to Thanh Ha from comment #10) > > hmm ok in that case since the "libwebkit-1.0.so.2" didn't exist until about > > an hour ago then we will need to get that onto the hudson-slaves too so that > > they all have this file. > > I spoke to Alexander on IRC and we decided to install this on all the slaves > so slaves 1-6 now have libwebkit-1.0.so.2 installed. Let me know if there's > anything else you need here. Thanks for installing the Webkit libraries on the hudson slaves Thanh! There seems to be a browser test failure (bug 426853) after this however, and I'm trying to confirm this in my local workspace. Can you let me know the exact version of the Webkit library that is installed on the hudson slaves? I'm guessing it is 1.2.x but I would like to confirm the same for my testing.
(In reply to David Williams from comment #12) > SWT: this sort of thing seems to be a recurring issue (such as from > community, those leading the bleeding edge with Ubuntu 13.10, etc.) I wonder > if we should have a clear statement of "what's required" (for both webkit > and xulrunner varieties) and ideally some directions for "how to tell what's > installed" since as far as I know, all the main distributions have different > "techniques" and repositories, and library naming conventions. (dpkg vs. > rpm, etc.). > > Anyway, I added "readme" to the keywords for that reason ... if you disagree > or have a different method of documenting that, feel free to change it. David, the requirements for running the SWT Browser on Linux are clearly mentioned in the SWT FAQ page here - http://eclipse.org/swt/faq.php#browserlinux We could add another entry about figuring out the installed packages and their versions etc. but IMHO that would not belong in the domain of Eclipse or SWT. Let me know what you think. Thanks!
(In reply to Arun Thondapu from comment #13) > Thanks for installing the Webkit libraries on the hudson slaves Thanh! There > seems to be a browser test failure (bug 426853) after this however, and I'm > trying to confirm this in my local workspace. Can you let me know the exact > version of the Webkit library that is installed on the hudson slaves? I'm > guessing it is 1.2.x but I would like to confirm the same for my testing. Arun, see comment 3 where I listed the exact versions installed. 1.2 is not available in SLES 11 unfortunately.
(In reply to Arun Thondapu from comment #14) > David, the requirements for running the SWT Browser on Linux are clearly > mentioned in the SWT FAQ page here - > http://eclipse.org/swt/faq.php#browserlinux > > We could add another entry about figuring out the installed packages and > their versions etc. but IMHO that would not belong in the domain of Eclipse > or SWT. Let me know what you think. Thanks! I see your point about not in the domain of Eclipse or SWT, and it is a good one. But, I feel if its not documented somewhere, we'll spend more time asking people what they have installed, and explaining "how to tell". The situation is similar to how we have "how to obtain a stack dump" page, at http://wiki.eclipse.org/How_to_report_a_deadlock We obviously can not cover all cases ... but ... we can document some cases, as they come up, and perhaps community would add to it. I don't know that much about "all distributions", but on the ones I use, I seldom see the types of names that FAQ refers to "Mozilla 1.4 GTK2 - 1.7.x GTK2, XULRunner 1.8.x - 1.9.x, 3.6.x, 10.x and 24.x (but not 2.x nor other unlisted versions), WebKitGTK+ 1.2.x and newer (Eclipse 4.4 and newer uses GTK 3". Maybe a compromise, that would help community, but "offload" the work from Eclipse and SWT Team is to have another wiki page (that the FAQ could point to) and on that wiki page make it clear this was as "community supported wiki page" so people could add their "tips and tricks" to determining what they have installed. I especially thinking of things like bug 424657 comment 10, where you not only need to obvious "webkit 3 runtime" installed, but also the "webkit 3 data". True, if some one installed the right browser, they might get both ... but ... it is sure hare to tell. If you don't object, I'll volunteer to get that wiki page started -- and we'll see ... it might not help ... but ... it might?
(In reply to Thanh Ha from comment #15) > (In reply to Arun Thondapu from comment #13) > > Thanks for installing the Webkit libraries on the hudson slaves Thanh! There > > seems to be a browser test failure (bug 426853) after this however, and I'm > > trying to confirm this in my local workspace. Can you let me know the exact > > version of the Webkit library that is installed on the hudson slaves? I'm > > guessing it is 1.2.x but I would like to confirm the same for my testing. > > Arun, see comment 3 where I listed the exact versions installed. 1.2 is not > available in SLES 11 unfortunately. Thanh, I'm interested in the version of the Webkit source that was used to generate these .so libraries and the versions of the .so file do not give me that information. For example, if you see this link - https://www.suse.com/LinuxPackages/packageRouter.jsp?product=desktop&version=11&service_pack=sp3&architecture=i386&package_name=libwebkit-1_0-2 - it lists the same libwebkit-1.0.so.2 library but the version of Webkit is 1.2.7 as specified at the top. The Webkit version installed on this SLES machine should also be 1.2.x if I'm not wrong and this is a supported version by SWT. I'll now need to confirm with the corresponding version of Webkit library in my local workspace and that is why I need the exact version of Webkit installed here.
(In reply to David Williams from comment #16) > > Maybe a compromise, that would help community, but "offload" the work from > Eclipse and SWT Team is to have another wiki page (that the FAQ could point > to) and on that wiki page make it clear this was as "community supported > wiki page" so people could add their "tips and tricks" to determining what > they have installed. I especially thinking of things like bug 424657 comment > 10, where you not only need to obvious "webkit 3 runtime" installed, but > also the "webkit 3 data". True, if some one installed the right browser, > they might get both ... but ... it is sure hare to tell. > > If you don't object, I'll volunteer to get that wiki page started -- and > we'll see ... it might not help ... but ... it might? Sure David, a community supported wiki page sounds like a very good idea to capture these distribution specific commands and other details. This should definitely help the wider community of users. I'll try to add whatever info I can once the page is ready. Thanks for taking it up :-)
(In reply to Arun Thondapu from comment #17) > Thanh, I'm interested in the version of the Webkit source that was used to > generate these .so libraries and the versions of the .so file do not give me > that information. For example, if you see this link - Arun, Sorry I had not realized there were additional versioning information. I ran the command "zypper info libwebkit-1_0-2" and I believe here is the result you're interested in: Information for package libwebkit-1_0-2: Name: libwebkit-1_0-2 Version: 1.2.7-0.15.2 Arch: x86_64 ...
Mind a novice question about "pre-reqs" ... Since I first heard about it, I thought our requirement for "WebKit" was/could be satisfied by having "Chome" installed. But, was just reading tonight "they" have their own fork of WebKit (they call WebKit Blink) ... so, I'm just curious ... can Eclipse users "get" their WebKit version from installing "Chrome"? Or must it be installed separately, much as they would a Java VM? I'm mostly asking for my own education, but can't help but wonder, could that be a reason why users seem to have such trouble "getting this working" ... because there are many similar, but slightly different versions of webkit around? I was casually researching this, after seeing that shortly after someone opened a serious "memory leak" bug, bug 321573, that a few hours later, by coincidence, Google released some fixes, some of which had to do with serious memory leaks related to images. Version 32.0.1700.102-0ubuntu0.12.04.1: * Upstream release 32.0.1700.102: - CVE-2013-6649: Use-after-free in SVG images. - CVE-2013-6650: Memory corruption in V8. This issue was fixed in v8 version 3.22.24.16. At first, I thought, "oh, maybe we need a webkit fix" ... but then as I learned more about it (admittedly, "on the internet" :) it sounds like they have a different web-kit ... I'm just curious how many of them are and can we keep them all straight! (And, I know the answer ... I know you can keep them all straight ... but hard to imagine how an end-user would. Again, no criticisms or complaints ... just trying to understand.
(In reply to David Williams from comment #20) > Mind a novice question about "pre-reqs" ... Since I first heard about it, I > thought our requirement for "WebKit" was/could be satisfied by having > "Chome" installed. But, was just reading tonight "they" have their own fork > of WebKit (they call WebKit Blink) ... so, I'm just curious ... can Eclipse > users "get" their WebKit version from installing "Chrome"? Or must it be > installed separately, much as they would a Java VM? I'm mostly asking for my > own education, but can't help but wonder, could that be a reason why users > seem to have such trouble "getting this working" ... because there are many > similar, but slightly different versions of webkit around? Eclipse on Linux works with webkitgtk only - http://webkitgtk.org/ . It can't work with QtWebKit or any of the other webkit variants. About Blink (Chromium engine) it's a fork not a variant of webkit aka the two projects have taken different roads AFAIK and the similarity between webkit and blink might not be bigger than between webkit and khtml soon. Also the JS implementation of webkit variants (JavaScriptCore) and the one of Chrome/Chromium/Blink (V8) are entirely different so even if there might be some similarities in the pure HTML rendering the JS integration would be entirely different. And yest webkitgtk is a prerequisite - it has to be installed on the user machine much in the same way OpenJDK and gtk have to be installed. The good part about webkitgtk is that it's part of default installation on many Linux distributions so it's the safest bet about being available on the user machine. Getting it from Chrome (or Chromium) would mean having additional Blink swt browser backend soon enough and working with it will require people to install Chrome/Chromium, which would be bigger request than installing webkitgtk. Now, the slightly different versions of webkit is caused by the fact that browser engine have new releases every 6 weeks (IIRC) while SWT tries to support all of these releases in the last 13 years - pretty much all webkitgtk releasses and selected xulrunner(mozilla). > I'm just curious how many of them are and can > we keep them all straight! (And, I know the answer ... I know you can keep > them all straight ... but hard to imagine how an end-user would. Keeping the pace with browser engines which release for the sake of browsers not for the sake of being embeddable rendering engine is more and more problematic and this seems to become the norm nowadays. I had few cases where I had to disable all SWT browser backends due to none of them being usable in given combination. There is no silver bullet in this area. In general being on still supported (by it's vendor not by SWT) Linux distribution gives good changes that it is still tested and potential engine breakages introduced are fixed. Hope that helps.
(In reply to Alexander Kurtakov from comment #21) > Hope that helps. It does. Clears up my misconception about being "tied to browsers" and its reassuring to hear "it's part of default installation on many Linux distributions". (Though, also makes me wonder which ones its not default ... and which future ones will change to an incompatible version -- guess we'll cross those bridges as we get to them.) Thank you,
(In reply to Thanh Ha from comment #19) > (In reply to Arun Thondapu from comment #17) > > Thanh, I'm interested in the version of the Webkit source that was used to > > generate these .so libraries and the versions of the .so file do not give me > > that information. For example, if you see this link - > > Arun, > > Sorry I had not realized there were additional versioning information. I ran > the command "zypper info libwebkit-1_0-2" and I believe here is the result > you're interested in: > > > Information for package libwebkit-1_0-2: > > Name: libwebkit-1_0-2 > Version: 1.2.7-0.15.2 > Arch: x86_64 > ... So, Arun, I assume this was the information you were looking for? Naive question ... does this version bring us up to "GTK 3", or merely "GTK 2"? And if merely 2, I'm assuming GTK 3 is "impossible" in SUSE 11? (Sorry if this was already covered ... but, I got lost and confused and thought I heard contradictory things).
(In reply to David Williams from comment #23) > > Information for package libwebkit-1_0-2: > > > > Name: libwebkit-1_0-2 > > Version: 1.2.7-0.15.2 > > Arch: x86_64 > > ... > > So, Arun, I assume this was the information you were looking for? Yes indeed. This was to verify the tests in my local workspace with the same version of Webkit as mentioned above (1.2.7), mainly to reproduce bug 426853. > Naive question ... does this version bring us up to "GTK 3", or merely "GTK > 2"? > And if merely 2, I'm assuming GTK 3 is "impossible" in SUSE 11? > (Sorry if this was already covered ... but, I got lost and confused and > thought I heard contradictory things). This is not connected to GTK 3 at all, this is still the GTK 2 version of Webkit that we're using and I'm reasonably certain that SLES 11 does not have official packages for GTK 3 or for Webkit's GTK 3 port. In fact, I doubt if it has packages available for higher versions of Webkit with GTK 2 itself (1.2.7 is a very old version, 1.2.x is the oldest SWT supports actually). Hope this makes things at least a bit clearer now...
Sorry to be a little off-topic, but I was thinking of starting that "community page" mentioned in previous comments and noticed something funny about my own installs ... Ubuntu 12.04.4 LTS (but, with custom kernel, which differs on my two machines in question: On one of my machines (laptop) I can run /usr/lib/webkitgtk-3.0-0/libexec/GtkLauncher http://acid3.acidtests.org and get some nice "debug?" output, which tells me that version of webkit is using GTK 2!? $ /usr/lib/webkitgtk-3.0-0/libexec/GtkLauncher http://acid3.acidtests.org No bp log location saved, using default. [000:000] Cpu: 6.42.7, x8, 2001Mhz, 16027MB [000:000] Computer model: Not available [000:000] Browser XEmbed support present: 1 [000:000] Browser toolkit is Gtk2. [000:002] Using Gtk2 toolkit No bp log location saved, using default. [000:000] Cpu: 6.42.7, x8, 2001Mhz, 16027MB [000:000] Computer model: Not available [000:036] No bp log location saved, using default. [000:037] Cpu: 6.42.7, x8, 2001Mhz, 16027MB [000:037] Computer model: Not available [000:037] Browser XEmbed support present: 1 [000:037] Browser toolkit is Gtk2. [000:037] Using Gtk2 toolkit [000:001] No bp log location saved, using default. [000:002] Cpu: 6.42.7, x8, 2001Mhz, 16027MB [000:002] Computer model: Not available On my other machine (desktop) I run same command, but get no "debug output" In both cases, the "acid test" displays correctly. If I use $ dpkg -l | grep "gtk.*-3" I have nearly identical "gtk 3 stuff" installed. Insights welcome ... but, obviously low priority. Does this mean Eclipse is using GTK2 on my systems? Or ... is that independent? Does this "make sense" to anyone? Or, is it simply a matter I'd have to build my own "webkit" for it to use GTK3?
(In reply to David Williams from comment #25) > Sorry to be a little off-topic, but I was thinking of starting that > "community page" mentioned in previous comments and noticed something funny > about my own installs ... Ubuntu 12.04.4 LTS (but, with custom kernel, which > differs on my two machines in question: > > On one of my machines (laptop) I can run > > /usr/lib/webkitgtk-3.0-0/libexec/GtkLauncher http://acid3.acidtests.org > > and get some nice "debug?" output, which tells me that version of webkit is > using GTK 2!? > > $ /usr/lib/webkitgtk-3.0-0/libexec/GtkLauncher http://acid3.acidtests.org > No bp log location saved, using default. > [000:000] Cpu: 6.42.7, x8, 2001Mhz, 16027MB > [000:000] Computer model: Not available > [000:000] Browser XEmbed support present: 1 > [000:000] Browser toolkit is Gtk2. > [000:002] Using Gtk2 toolkit > No bp log location saved, using default. > [000:000] Cpu: 6.42.7, x8, 2001Mhz, 16027MB > [000:000] Computer model: Not available > [000:036] No bp log location saved, using default. > [000:037] Cpu: 6.42.7, x8, 2001Mhz, 16027MB > [000:037] Computer model: Not available > [000:037] Browser XEmbed support present: 1 > [000:037] Browser toolkit is Gtk2. > [000:037] Using Gtk2 toolkit > [000:001] No bp log location saved, using default. > [000:002] Cpu: 6.42.7, x8, 2001Mhz, 16027MB > [000:002] Computer model: Not available > > > On my other machine (desktop) I run same command, but get no "debug output" > > In both cases, the "acid test" displays correctly. > > If I use > $ dpkg -l | grep "gtk.*-3" > I have nearly identical "gtk 3 stuff" installed. > > Insights welcome ... but, obviously low priority. > > Does this mean Eclipse is using GTK2 on my systems? > Or ... is that independent? > > Does this "make sense" to anyone? Or, is it simply a matter I'd have to > build my own "webkit" for it to use GTK3? Hi David, sorry I missed your comment for so long. These gtk2 lines are caused by browser plugins that haven't been migrated yet. webkitgtk3 and eclipse can run on GTK3 but browser plugins can still use gtk2 assuming they run the gtk2 stuff out of process otherwise it is a recipe for disaster as we have seen with google talk and bluejeans plugins before they were fixed.
afaik we're moving to webkit2 and mozlla's backend is not being updated anymore? Should we close this bug?
(In reply to Leo Ufimtsev from comment #27) > afaik we're moving to webkit2 and mozlla's backend is not being updated > anymore? Should we close this bug? Yes, irrelevant now that mozilla backend is dead.
