Bugzilla – Full Text Bug Listing
| Summary: | [GTK3] [32-bit] NoSuchMethodError: gObjectClass_finalize | ||
|---|---|---|---|
| Product: | [Eclipse Project] Platform | Reporter: | Thomas Singer <eclipse> |
| Component: | SWT | Assignee: | Platform-SWT-Inbox <platform-swt-inbox> |
| Status: | CLOSED WONTFIX | QA Contact: | |
| Severity: | blocker | ||
| Priority: | P3 | CC: | akurtakov, daniel_megert, ericwill, Lars.Vogel, loskutov, peter, s.muecke, simeon.danailov.andreev |
| Version: | 4.8 | Keywords: | triaged |
| Target Milestone: | --- | ||
| Hardware: | PC | ||
| OS: | Linux | ||
| See Also: |
https://bugs.eclipse.org/bugs/show_bug.cgi?id=470031 https://git.eclipse.org/r/112104 https://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=28cccc6c31636d2dfce20e6fd2b67b306db0ddbb https://bugs.eclipse.org/bugs/show_bug.cgi?id=536764 https://bugs.eclipse.org/bugs/show_bug.cgi?id=536766 |
||
| Whiteboard: | |||
|
Description
Thomas Singer
(In reply to Thomas Singer from comment #0) > Disposing a shell with the latest SWT 4.830 causes following exception to > occur on Linux Mint 18.1 with GTK 3.18.9 and Mint-X theme: > > java.lang.NoSuchMethodError: gObjectClass_finalize > at org.eclipse.swt.internal.gtk.OS._gtk_widget_destroy(Native Method) > at org.eclipse.swt.internal.gtk.OS.gtk_widget_destroy(OS.java:13691) > at org.eclipse.swt.widgets.Widget.destroyWidget(Widget.java:463) > at org.eclipse.swt.widgets.Widget.release(Widget.java:1186) > at org.eclipse.swt.widgets.Control.release(Control.java:4117) > at org.eclipse.swt.widgets.Widget.dispose(Widget.java:498) > at org.eclipse.swt.widgets.Shell.dispose(Shell.java:2700) Is this with a fresh install? Can you try with http://download.eclipse.org/eclipse/downloads/drops4/S-4.8M3a-201710300400/ I've tried with today's master and Snippet1.java, no issues. (gtk3.22/Fedora26). Please kindly post more details: - Snippet to reproduce (or link to SWT Snippets snippet) - Last known good version(?) (roughly). Maybe it's trying to load non-existing bindings? Try to move ~/.swt to allow new .so files to be extracted. @Eric, any ideas? gObjectClass_finalize() is a method in AccessibleObject.java that is called via JNI from os_custom.c. This has been the case for ~2 months now, as part of the new accessibility work. It looks like JNI can't find the method, which I'm guessing means something is wrong with your binaries/bindings. Maybe try a different build? I can't reproduce the issue on GTK3.18 (or 3.22). (In reply to Dani Megert from comment #1) > (In reply to Thomas Singer from comment #0) > > Disposing a shell with the latest SWT 4.830 causes following exception to > > occur on Linux Mint 18.1 with GTK 3.18.9 and Mint-X theme: > > > > java.lang.NoSuchMethodError: gObjectClass_finalize > > at org.eclipse.swt.internal.gtk.OS._gtk_widget_destroy(Native Method) > > at org.eclipse.swt.internal.gtk.OS.gtk_widget_destroy(OS.java:13691) > > at org.eclipse.swt.widgets.Widget.destroyWidget(Widget.java:463) > > at org.eclipse.swt.widgets.Widget.release(Widget.java:1186) > > at org.eclipse.swt.widgets.Control.release(Control.java:4117) > > at org.eclipse.swt.widgets.Widget.dispose(Widget.java:498) > > at org.eclipse.swt.widgets.Shell.dispose(Shell.java:2700) > > Is this with a fresh install? Can you try with > http://download.eclipse.org/eclipse/downloads/drops4/S-4.8M3a-201710300400/ It is a fresh build from sources - like I do it since approx. 3 or 4 years. I reckon, it wouldn't work with SWT_GTK3=0 if the build would have failed. Does this happen all the time? With a specific snippet? (In reply to Thomas Singer from comment #4) > (In reply to Dani Megert from comment #1) > > (In reply to Thomas Singer from comment #0) > > > Disposing a shell with the latest SWT 4.830 causes following exception to > > > occur on Linux Mint 18.1 with GTK 3.18.9 and Mint-X theme: > > > > > > java.lang.NoSuchMethodError: gObjectClass_finalize > > > at org.eclipse.swt.internal.gtk.OS._gtk_widget_destroy(Native Method) > > > at org.eclipse.swt.internal.gtk.OS.gtk_widget_destroy(OS.java:13691) > > > at org.eclipse.swt.widgets.Widget.destroyWidget(Widget.java:463) > > > at org.eclipse.swt.widgets.Widget.release(Widget.java:1186) > > > at org.eclipse.swt.widgets.Control.release(Control.java:4117) > > > at org.eclipse.swt.widgets.Widget.dispose(Widget.java:498) > > > at org.eclipse.swt.widgets.Shell.dispose(Shell.java:2700) > > > > Is this with a fresh install? Can you try with > > http://download.eclipse.org/eclipse/downloads/drops4/S-4.8M3a-201710300400/ > > It is a fresh build from sources - like I do it since approx. 3 or 4 years. Can you please try with the suggested build. I have performed further tests. Steps: - download SmartSynchronize 3.4.12 from <http://www.syntevo.com/smartsynchronize/download?file=smartsynchronize/smartsynchronize-linux-3_4_12.tar.gz> - unpack bin/smartsynchronize.sh, run it from command line until the "Welcome" window occurs - press Alt+X (or click the "Exit" button) To eliminate anything from our launch-script (bin/smartsynchronize.sh), one could launch it also from the lib directory using following command: $ env SWT_GTK3=1 java -jar bootloader.jar The log at ~/.smartsynchronize/log.txt should indicate what jars actually are loaded. I have replaced lib/org.eclipse.swt.gtk.linux.x86.jar with the swt.jar from http://download.eclipse.org/eclipse/downloads/drops4/S-4.8M3a-201710300400/ - same crash. I have bisected the SWT sources and found out that the bug has been introduced with build 4.820. The last working commit was 3154c28ad77 "Bug 522718 - Remove mozilla leftovers" from 2017-09-25 12:39 and the first bad (successfully compilable) commit being 7448b63396e "v4820" from 2017-09-26 10:10 To me it looks like the commit which introduced the crash was bdaf68d5b28 "Bug 470031: [GTK3] Some viewparts became inaccessible in 4.5" from 2017-09-06 17:47 The bad thing is, I could not reproduce it within a snippet. I was not even possible to reproduce it when launching SmartSynchronize from within IDEA. (In reply to Thomas Singer from comment #8) > - unpack bin/smartsynchronize.sh, run it from command line until the > "Welcome" window occurs should read - unpack, - run bin/smartsynchronize.sh from command line until the "Welcome" window occurs I will investigate. (In reply to Thomas Singer from comment #8) > I have replaced lib/org.eclipse.swt.gtk.linux.x86.jar with the swt.jar from > ^- 32 bit? Are you using a 32bit of swt? Theory: Some pointer issue where 32/64 bit pointers might got muddled up. (who knows??) Are you able to test with a 64 bit version? Btw, may be of interest if using 32 bit: Bug 526620 – Future of 32-bit Eclipse https://bugs.eclipse.org/bugs/show_bug.cgi?id=526620 There are plans on dropping 32 bit version. If you depend on 32bit, consider hoping onto the 64bit wagon if possible. Yes, I'm using the 32-bit lib because I have a 32-bit Linux installed. (In reply to Thomas Singer from comment #8) > I have performed further tests. Steps: > - download SmartSynchronize 3.4.12 from > <http://www.syntevo.com/smartsynchronize/download?file=smartsynchronize/ > smartsynchronize-linux-3_4_12.tar.gz> > - unpack bin/smartsynchronize.sh, run it from command line until the > "Welcome" window occurs > - press Alt+X (or click the "Exit" button) > > To eliminate anything from our launch-script (bin/smartsynchronize.sh), one > could launch it also from the lib directory using following command: > > $ env SWT_GTK3=1 java -jar bootloader.jar > > The log at ~/.smartsynchronize/log.txt should indicate what jars actually > are loaded. > > I have replaced lib/org.eclipse.swt.gtk.linux.x86.jar with the swt.jar from > http://download.eclipse.org/eclipse/downloads/drops4/S-4.8M3a-201710300400/ > - same crash. I have bisected the SWT sources and found out that the bug has > been introduced with build 4.820. The last working commit was > > 3154c28ad77 "Bug 522718 - Remove mozilla leftovers" from 2017-09-25 12:39 > > and the first bad (successfully compilable) commit being > > 7448b63396e "v4820" from 2017-09-26 10:10 > > To me it looks like the commit which introduced the crash was > > bdaf68d5b28 "Bug 470031: [GTK3] Some viewparts became inaccessible in 4.5" > from 2017-09-06 17:47 > > > The bad thing is, I could not reproduce it within a snippet. I was not even > possible to reproduce it when launching SmartSynchronize from within IDEA. I did these steps and could not get the issue to reproduce. This is on a 64-bit OS, using the 64-bit SWT libs. I will try on a 32-bit VM to see if the issue reproduces...even if it doesn't, I believe I'll be able to patch it since it's a JNI error which I can catch. (In reply to Thomas Singer from comment #12) > Yes, I'm using the 32-bit lib because I have a 32-bit Linux installed. Out of technical curiosity, is it possible for you guys to migrate to 64 bit, or would it currently be too expensive/time consuming? (I'm kinda thinking about what to do when 32bit will be dropped...). (In reply to Leo Ufimtsev from comment #14) > (In reply to Thomas Singer from comment #12) > > Yes, I'm using the 32-bit lib because I have a 32-bit Linux installed. > > Out of technical curiosity, is it possible for you guys to migrate to 64 > bit, or would it currently be too expensive/time consuming? (I'm kinda > thinking about what to do when 32bit will be dropped...). Not saying that we should ignore 32 bit bugs :-), just curious. (In reply to Leo Ufimtsev from comment #15) > (In reply to Leo Ufimtsev from comment #14) > > (In reply to Thomas Singer from comment #12) > > > Yes, I'm using the 32-bit lib because I have a 32-bit Linux installed. > > > > Out of technical curiosity, is it possible for you guys to migrate to 64 > > bit, or would it currently be too expensive/time consuming? (I'm kinda > > thinking about what to do when 32bit will be dropped...). > > Not saying that we should ignore 32 bit bugs :-), just curious. See bug 526620. (In reply to Leo Ufimtsev from comment #14) > (In reply to Thomas Singer from comment #12) > > Yes, I'm using the 32-bit lib because I have a 32-bit Linux installed. > > Out of technical curiosity, is it possible for you guys to migrate to 64 > bit, or would it currently be too expensive/time consuming? (I'm kinda > thinking about what to do when 32bit will be dropped...). Sure, this would be possible, but ATM our Linux test machine just has a 32-bit Linux installed. Okay, I can reproduce the issue (kind of) in my VM. The crash/error doesn't actually happen, but I see the error message: CRITICAL **: JNI method ID pointer is NULL for method gObjectClass_finalize Which is undoubtedly causing the crash on your end. I can prepare a patch, Thomas would be able to review + test it? (In reply to Eric Williams from comment #18) > Okay, I can reproduce the issue (kind of) in my VM. > > The crash/error doesn't actually happen, but I see the error message: > > CRITICAL **: JNI method ID pointer is NULL for method gObjectClass_finalize > > Which is undoubtedly causing the crash on your end. I can prepare a patch, > Thomas would be able to review + test it? Yes, this should be possible tomorrow. New Gerrit change created: https://git.eclipse.org/r/112104 Sorry, I can't build native code and hence can't verify the patch. (In reply to Thomas Singer from comment #21) > Sorry, I can't build native code and hence can't verify the patch. Okay. I will build it myself and try on my VM. If it fixes the errors I'll push it. You can check out the I-build tomorrow and verify on your machine. Gerrit change https://git.eclipse.org/r/112104 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=28cccc6c31636d2dfce20e6fd2b67b306db0ddbb (In reply to Eclipse Genie from comment #23) > Gerrit change https://git.eclipse.org/r/112104 was merged to [master]. > Commit: > http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/ > ?id=28cccc6c31636d2dfce20e6fd2b67b306db0ddbb I've merged the patch. Building SWT on 32-bit required a lot of time so I didn't go down that road. Please test with tomorrow's I-build and let me know if the issue is fixed. I'll keep this bug open in the meantime. Sorry, Eric, the bug is still present in 4.832. I'm also still seeing the output CRITICAL **: JNI method ID pointer is NULL for method gObjectClass_finalize printed to stderr. (In reply to Thomas Singer from comment #25) > Sorry, Eric, the bug is still present in 4.832. I'm also still seeing the > output > > CRITICAL **: JNI method ID pointer is NULL for method gObjectClass_finalize > > printed to stderr. Alright, I'll investigate further next week. What is the status of this bug guys? (In reply to Alexander Kurtakov from comment #27) > What is the status of this bug guys? I still cannot reproduce the bug. Should we close this one? The crash still is reproducible with stock Linux Mint 18.3 Cinnamon/32-bit. See also bug 536764 and bug 536766. 32 bit Linux support is being dropped in 4.10. See bug bug 536766. *** Bug 536764 has been marked as a duplicate of this bug. *** |