815025 - gcc 4.2.1 build: nsDOMFile.o: Assembler messages: Error: junk at end of line, first unrecognized character is `*'

Status: RESOLVED WONTFIX

(Core :: MFBT, defect)

Description

[Jan Beich]

Attachment: workaround

When trying to build firefox 18.0b1 with /usr/bin/g++ on FreeBSD it fails with:

  $ gmake
  [...]
  {standard input}: Assembler messages:
  {standard input}:7643: Error: junk at end of line, first unrecognized character is `*'
  gmake[5]: *** [nsDOMFile.o] Error 1

  $ g++ -v
  Using built-in specs.
  Target: amd64-undermydesk-freebsd
  Configured with: FreeBSD/amd64 system compiler
  Thread model: posix
  gcc version 4.2.1 20070831 patched [FreeBSD]

The issue seems to be with SHA1Sum constructor declared as weak i.e., GCC bug c++/36107. So, do BSDs need another autoconf test like in bug 788276?

fwiw, I use few more workarounds (like bug 732340, bug 783505, bug 798354) to make build succeed.

Comment 1

[Landry Breuil (:gaston)]

I dont know how you got there, since here on mozilla-central (and i think mozilla-beta) configure doesnt pass if CC is set to use openbsd's gcc 4.2.1.

checking For gcc visibility bug with class-level attributes (GCC bug 26905)... no
checking For x86_64 gcc visibility bug with builtins (GCC bug 20297)... no
checking for gcc PR49911... no
checking for gcc pr39608... yes
This compiler would fail to build firefox, plase upgrade.

(note the typo :)

Comment 2

[Rafael Ávila de Espíndola (:espindola) (not reading bugmail)]

A patch fixing the typo is OK :-)

I really don't think we should put effort into making making the build work with gcc 4.2.

Comment 3

[Joshua Cranmer [:jcranmer]]

(In reply to Rafael Ávila de Espíndola (:espindola) from comment #2)
> A patch fixing the typo is OK :-)
> 
> I really don't think we should put effort into making making the build work
> with gcc 4.2.

Our minimum build requirement is now gcc 4.4, and I suspect we'd move to 4.5 if/when we upgrade compilers on android builds.

Comment 4

[Jeff Walden [:Waldo]]

Sounds WONTFIXish to me given the compiler requirement cited in comment 3.

Comment 5

[Landry Breuil (:gaston)]

WONTFIX for gcc 4.2 compat yes, but maybe we should try to understand how 18.0b1 configure did pass with gcc 4.2.1 on FreeBSD, whereas it should have errored out. I've just retested 18.0b1 on OpenBSD, and configure doesnt pass with 4.2.1.

Comment 6

[Jeff Walden [:Waldo]]

Fine enough if you want to check that, but please do so in a separate bug, targeted to that specific issue.

Comment 7

[Jan Beich]

Do you mind if I file a few more such workarounds as WONTFIX to document what we have in ports? FreeBSD has quite a few users still clinging to ancient GCC (patched but still under GPLv2) in base while Clang switch happened only recently (in 2012-11-05).

(In reply to Landry Breuil (:gaston) from comment #5)
> WONTFIX for gcc 4.2 compat yes, but maybe we should try to understand how
> 18.0b1 configure did pass with gcc 4.2.1 on FreeBSD, whereas it should have
> errored out. I've just retested 18.0b1 on OpenBSD, and configure doesnt pass
> with 4.2.1.

http://trillian.chruetertee.ch/freebsd-gecko/browser/trunk/www/firefox-nightly/files/patch-bug783505