Created attachment 493498 [details] compressed build.log I tried to build firefox on an arm chroot with these useflags [ebuild N ~] www-client/firefox-52.3.0::gentoo USE="dbus gtk2 jemalloc system-harfbuzz system-icu system-jpeg system-libevent system-sqlite -bindist -custom-cflags -custom-optimization -debug -gmp-autoupdate (-hardened) -hwaccel -jack -neon (-pgo) -pulseaudio (-rust) (-selinux) -startup-notification (-system-cairo) -system-libvpx {-test} -wifi" L10N="-ach -af -an -ar -as -ast -az -bg -bn-BD -bn-IN -br -bs -ca -cak -cs -cy -da -de -dsb -el -en-GB -en-ZA -eo -es-AR -es-CL -es-ES -es-MX -et -eu -fa -ff -fi -fr -fy -ga -gd -gl -gn -gu -he -hi -hr -hsb -hu -hy -id -is -it -ja -ka -kab -kk -km -kn -ko -lij -lt -lv -mai -mk -ml -mr -ms -nb -nl -nn -or -pa -pl -pt-BR -pt-PT -rm -ro -ru -si -sk -sl -son -sq -sr -sv -ta -te -th -tr -uk -uz -vi -xh -zh-CN -zh-TW" 0 KiB Total: 1 package (1 reinstall), Size of downloads: 0 KiB The build fails, here is a snipet of build log (full build log attached) [01m[K/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/ipc/contentproc/plugin-container.cpp:[m[K In function '[01m[Kint content_process_main(int, char**)[m[K': [01m[K/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/ipc/contentproc/plugin-container.cpp:168:5:[m[K [01;31m[Kerror: [m[K'[01m[KSandboxEarlyInit[m[K' is not a member of '[01m[Kmozilla[m[K' mozilla::SandboxEarlyInit(XRE_GetProcessType()); [01;32m[K ^[m[K make[4]: *** [/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/config/rules.mk:953: plugin-container.o] Error 1 make[4]: Leaving directory '/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/ff/ipc/contentproc' make[3]: *** [/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/config/recurse.mk:71: ipc/contentproc/target] Error 2 make[3]: Leaving directory '/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/ff' make[2]: *** [/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/config/recurse.mk:33: compile] Error 2 make[2]: Leaving directory '/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/ff' make[1]: *** [/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/config/rules.mk:523: default] Error 2 make[1]: Leaving directory '/var/tmp/portage/www-client/firefox-52.3.0/work/firefox-52.3.0esr/ff' make: *** [client.mk:415: realbuild] Error 2
Created attachment 493500 [details] output of emerge --info I dropped MAKEOPTS to -j1 via /etc/portage/env/
This is the relevant part of the source code, in case someone would like to check it: https://github.com/mozilla/gecko-dev/blob/esr52/ipc/contentproc/plugin-container.cpp#L168
(In reply to tt_1 from comment #2) > This is the relevant part of the source code, in case someone would like to > check it: > > https://github.com/mozilla/gecko-dev/blob/esr52/ipc/contentproc/plugin- > container.cpp#L168 No point in checking it, I am gonna go ahead and disable the sandbox again in esr build.
Can you give me two to three days to elaborate this further? I used the patched eclass and could build successfully, but also I could get past the compilation of the plugin-container by using --enable-sandbox and --disable-content-sandbox (the compile of this build is still running) Also there must have been a reason from your point of view to activate them in the first place?
Sandbox is disabled on esr builds, change made via eclass so no revision bumps
Thanks for the fix, to append --disable-content-sandbox would have had been enough to fix this. For your information only and in case you plan to reenable it on 52.x.esr series.
Shouldn't this bug be sent upstream or resolved with a specific gentoo patch for this version to really resolve the problem instead of a workaround? I have investigated and was just ready, with all files prepared, to send another bug report of this same thing. I had this problem with FF 52.2.0 and I found that the file ${PORTAGE_TMPDIR}/portage/www-client/firefox-52.2.0/work/firefox-52.2.0esr/ipc/contentproc/plugin-container.cpp may have a bug with compiler directives: - There are several tests "#if defined(XP_LINUX) && defined(MOZ_GMP_SANDBOX)"; - Only on the line 164 there is a test "#if defined(XP_LINUX) && defined(MOZ_SANDBOX)"; - that may be mixed with the "#if defined(XP_WIN) && defined(MOZ_SANDBOX)" also present several times. This file uses a function from an include that is not used as the directives test gives false on the include area and true on the snippet that the function is called. I searched the online code of Firefox, trunk, and this same file is smaller, there is no test for XP_LINUX, only XP_WIN, so my comment will not apply there, it just apply if them maintain older versions too.
Forgot to say that Firefox compiles properly when the line 164 is changed to match the tests on other lines. Sorry for not telling this on my prior comment. :-[
Hey, I think you are the most welcome to solve this issue with a patch. Unfortunately I am not really capable of understanding all these includes and c/c++ in general, but it might be worth a try to change them accordingly. Please provide a patch, I can test it including runtime testing. Also please note that the compile passes for amd64 but fails for arm with --enable-content-sandbox, and that the comment of the source code mentions something about Android. It might be that you could end up with a more elegant solution, such as a defined testing to differentiate between arm_linux and arm_android.
Ok, just give me two days to learn how to include a patch on the portage structure (no more I hope) and another day to test compilation of the 52.3.0 on my Raspberry Pi 2.
The firefox ebuild has eapply_user, all you have to do is to copy your patch into /etc/portage/patches/www-client/firefox or www-client/firefox-52.3.0 if you want to be more specific. src_config for firefox can be altered via a simple txtfile /etc/portage/env/www-client/firefox , use EXTRA_ECONF="" and the command you'd like to overwrite. No need to edit the mozilla eclass.
Created attachment 495714 [details, diff] Fix for a compiler directives test on file plugin-container.cpp, for firefox 52.3.0
Created attachment 495716 [details, diff] Fix for a compiler directives test on file plugin-container.cpp, for firefox 52.2.0
Thanks to you, no need for the first two days! :-D It is compiling now and I hope to get a nice compiled FF in about 8 hours. :-) So, if you want to try before my test completes (I think may be the case based on the way You wrote the comment #9 ;-) ), I have included two patches, one for the 52.2.0 and another for the 52.3.0, I did as you mentioned on comment #11 and put the patches on /etc/portage/patches/www-client/firefox-52.x.x, changing x for the right numbers for each version. They are version specific because I don't if it is possible to use the same file for both versions, as the top of the patches there is a path for the file and the first directory includes the version of the package. Just to clarify in case of any doubt: it is not a full blind patch like "everything is like that and only this is different so it must be wrong", it is obscure, but I searched at least some clues. I wanted to sumarize on my first comment #7, and may be I did it too mutch. I searched where the function of the offender line (168) were defined, because it is from a namespace called "mozilla" so it should be on the sources being compiled. I found it on the file "mozilla/Sandbox.h" or "mozilla/SandboxInfo.h", I don't remember. On the plugin-container.cpp file these are included on the lines 30 and 31. Well, for these includes take place, the compiler directive "#if defined(XP_LINUX) && defined(MOZ_GMP_SANDBOX)" have to be true, but the function on line 168 is run when the compiler directive on line 164 is true, and that is "#if defined(XP_LINUX) && defined(MOZ_SANDBOX)". They are not equal, of course there will be a problem, the code uses a function that is on an include that can be excluded by the directives. Only then I realized that all other tests for with the "XP_LINUX" had the "GMP". I plan to tell the compilation results of the 52.3.0 at 11:00 to 12:00 UTC, about 10 hours from now.
I am writing this from the new compilation of www-client/firefox-52.3.0! :-) Compilation finished in 32997s, that is 9h9m57s. (RPi2, -j2, 1GB RAM and 1G swap) Now I agree that the issue is solved! :-) May be the attention from the maintainer is needed now.
Yes, indeed, the patch fixes the compilation for arm and --enable-sandbox and content sandbox. Runtime testing done, no issues so fare. It would be really helpfull if you can open a bug upstream at mozillas bugtracker and explain to them why they should accept this patch. Support for arm is not seen as a prioritiy for them, hence they will not find and fix this bug on their own. But they may know why this breaks arm and not x86_64, and if there are any further issues with the code regarding android (could imagine that while reading the comment in the source code next to the fixed lines) I don't think the gentoo mozilla team will reenable the config options in the eclass because of https://bugs.gentoo.org/627426
Bug sent upstream: https://bugzilla.mozilla.org/show_bug.cgi?id=1402133
I wrote you an email with some more informations and questions, please check your inbox.
