| Summary: | mail-client/thunderbird-60.6.1 fails at start of install phase | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Daniel Santos <daniel.santos> |
| Component: | Current packages | Assignee: | Mozilla Gentoo Team <mozilla> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | normal | CC: | adpc, jstein |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
xz'd build log
emerge --info |
||
|
Description
Daniel Santos
2019-05-25 21:11:20 UTC
Jonas Stein: Can't attach 1452KiB gzipped log file: Request Timeout Server timeout waiting for the HTTP request from the client. Apache Server at bugs.gentoo.org Port 443 Created attachment 577800 [details]
xz'd build log
Attempting xz -9e of log file...
Created attachment 577802 [details]
emerge --info
This is my emerge --info mail-client/thunderbird after I hacked the install to complete. Of note, libxul.so was 2.1GiB which is causing me to suspect some 32-bit signed file size issue -- but again it doesn't happen when I run elfhack as a normal user. HOWEVER, when I did so, I found no difference between the resulting file and a copy of the original file, so it could have just failed silently. I'm not actually certain what their hackelf is supposed to be for.
I've filed an upstream report: https://bugzilla.mozilla.org/show_bug.cgi?id=1554428 This is almost certainly due to using 32 bit values for offsets, even though it initially appears to be unsigned int, it's probably treated as signed elsewhere and ELF64 uses 64-bit offsets. build/unix/elfhack/elfxx.h:284: ElfSection *getSectionAt(unsigned int offset); The work-around is almost certainly to build without -g*. I can confirm this bug. The option "-g" seems to cause this bug. *** Bug 686644 has been marked as a duplicate of this bug. *** (In reply to ad PC from comment #5) > I can confirm this bug. The option "-g" seems to cause this bug. Thank you ad PC! So I propose we configure with --disable-elf-hack if -g* is in CFLAGS or CXXFLAGS. (In reply to Daniel Santos from comment #7) > (In reply to ad PC from comment #5) > > I can confirm this bug. The option "-g" seems to cause this bug. > > Thank you ad PC! So I propose we configure with --disable-elf-hack if -g* > is in CFLAGS or CXXFLAGS. what is to propose really, I have addressed the issue already. (In reply to Jory A. Pratt from comment #8) > (In reply to Daniel Santos from comment #7) > > (In reply to ad PC from comment #5) > > > I can confirm this bug. The option "-g" seems to cause this bug. > > > > Thank you ad PC! So I propose we configure with --disable-elf-hack if -g* > > is in CFLAGS or CXXFLAGS. > > what is to propose really, I have addressed the issue already. A comment in a bug report is not a fix. It needs a patch to the ebuild that prevents the problem from occurring. Also, upstream needs to fix their elfhack code to use appropriate data types. They're already in C++, there's no reason to not detect if their 32 or 64 bit and use the correct templatized types, or just typedef them and build support for both elf32 and elf64. *** This bug has been marked as a duplicate of bug 681438 *** @Jory: The proposed patch does not work for me. I have only "-g" in CFLAGS and CXXFLAGS and not "-ggdb3". |