237487 - SVG-enabled FireFox fails to build due missing |alloca()| symbol

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Roland Mainz]

An attempt to build a SVG-enabled FireFox fails due missing |alloca()| symbol on
Solaris2.8/SPARC with Sun Workshop 8.

The failure occurs at the final link:
-- snip --
/opt/SUNWspro/bin/CC -o firefox-bin -I/usr/openwin/include -features=tmplife
-xbuiltin=%all -mt  -DNDEBUG -DTRIMMED -xO2  nsBrowserApp.o
nsStaticComponents.o  -xildoff -zlazyload -zcombreloc    -L../../dist/bin
-L../../dist/lib -L../../dist/lib/components -lxpcom_compat_c -lxpconnect
-luconv -lucvmath -li18n -lctl -loji -lnecko -lnecko2 -lpref -lcaps -lrdf
-lhtmlpars -lgfxxprint -lgfx_gtk -limglib2 -lgkplugin -lwidget_gtk
-lxremote_client -lgklayout -lmork -ldocshell -lprofile -lnsprefm
-lembedcomponents -lwebbrwsr -leditor -ltxmgr -lcomposer -lnsappshell -lchrome
-laccessibility -lpipboot -lpipnss -lpippki -lfileview -lmozfind
-lxremoteservice -lappcomps -ltoolkitcomps -lcookie -lxmlextras -lautoconfig
-ltransformiix -luniversalchardet -ltypeaheadfind -lwebsrvcs -lnkfinger
-lnkdatetime -lbrowsercomps -lunicharutil_s -lucvutil_s -lmozpango -ljsj
-lgtksuperwin -lgtkxtbin -lgfxshared_s -lxlibrgb -lxprintutil -lgkgfx -lxulapp_s
../../dist/lib/libxulapp_s.a -L../../dist/bin -lmozjs -L../../dist/bin -lxpcom 
-L/shared/bigtmp/mozilla/nightlybuilds/north/objdir_gtk1/dist/lib -lplds4 -lplc4
-lnspr4 -lpthread -ldl -lrt  -L/usr/local/lib -L/usr/openwin/lib
-R/usr/openwin/lib -lgtk -lgdk -lgmodule -lglib -ldl -lXext -lX11 -lsocket -lnsl
-lm  -lsocket -ldl -lm  -L../../dist/lib/components -L../../dist/lib -lmozpng
-L../../dist/lib -lmozjpeg -L../../dist/lib -lmozz  -L../../dist/bin
-L../../dist/lib ../../dist/lib/libcrmf.a -lsmime3 -lssl3 -lnss3 -lsoftokn3  
-lmoz_art_lgpl -lm  -L/usr/openwin/lib -R/usr/openwin/lib -lXt
-L/usr/openwin/lib -R/usr/openwin/lib -lXp -lXext -lX11 -L../../dist/lib
-lxpcom_compat    
Undefined                       first referenced
 symbol                             in file
alloca                              ../../dist/bin/libmoz_art_lgpl.so
ld: fatal: Symbol referencing errors. No output written to firefox-bin
gmake[3]: *** [firefox-bin] Error 1
gmake[3]: Leaving directory
`/shared/bigtmp/mozilla/nightlybuilds/north/objdir_gtk1/browser/app'
gmake[2]: *** [libs] Error 2
gmake[2]: Leaving directory
`/shared/bigtmp/mozilla/nightlybuilds/north/objdir_gtk1/browser'
gmake[1]: *** [tier_99] Error 2
gmake[1]: Leaving directory
`/shared/bigtmp/mozilla/nightlybuilds/north/objdir_gtk1'
gmake: *** [default] Error 2
-- snip --

Comment 1

[Boris Zbarsky [:bzbarsky]]

Note that the alloca manpage has the following to say:

CONFORMING TO
       There is evidence that the alloca function appeared in 32v, pwb, pwb.2,
       3bsd, and 4bsd.  There is a man page for it in BSD 4.3.  Linux uses the
       GNU version.  This function is not in POSIX or SUSv3.

BUGS
       The  alloca function is machine and compiler dependent. On many systems
       its implementation is buggy. Its use is discouraged.

It looks like the problem is that the code at
http://lxr.mozilla.org/mozilla/source/other-licenses/libart_lgpl/art_render_gradient.c#35
will declare the function on your platform (which presumably does not define
HAS_ALLOCA_H) but will not implement it....

Comment 2

[Roland Mainz]

Talking myself... I have a fix for that...

Comment 3

[Roland Mainz]

Attachment: Patch for 2004-03-14-trunk

Comment 4

[Roland Mainz]

Comment on attachment 143919 [details] [diff] [review]
Patch for 2004-03-14-trunk

Requesting r= and checkin (pacth not part of default build).

Comment 5

[Boris Zbarsky [:bzbarsky]]

shouldn't we have a configure test for alloca.h?  Something like
http://lxr.mozilla.org/mozilla/source/js2/configure#3180 ?

Comment 6

[Alex Fritze]

Comment on attachment 143919 [details] [diff] [review]
Patch for 2004-03-14-trunk

Boris is right. Rather than patching libart, we need 
'AC_FUNC_ALLOCA' in configure.in so that HAVE_ALLOCA_H is set correctly.

Comment 7

[Roland Mainz]

(In reply to comment #6)
> (From update of attachment 143919 [details] [diff] [review])
> Boris is right. Rather than patching libart, we need 
> 'AC_FUNC_ALLOCA' in configure.in so that HAVE_ALLOCA_H is set correctly.

Why has AIX then extra rules ? Or HP-UX ?

Reassining bug back to default over... better I don't touch configure.in, I
always cause bustage when I touch the build system.

Comment 8

[Alex Fritze]

(In reply to comment #7)
> Why has AIX then extra rules ? Or HP-UX ?

From the autoconf manual:

     Source files that use `alloca' should start with a piece of code
     like the following, to declare it properly.  In some versions of
     AIX, the declaration of `alloca' must precede everything else
     except for comments and preprocessor directives.  The `#pragma'
     directive is indented so that pre-ANSI C compilers will ignore it,
     rather than choke on it.

          /* AIX requires this to be the first thing in the file.  */
          #ifndef __GNUC__
          # if HAVE_ALLOCA_H
          #  include <alloca.h>
          # else
          #  ifdef _AIX
           #pragma alloca
          #  else
          #   ifndef alloca /* predefined by HP cc +Olibcalls */
          char *alloca ();
          #   endif
          #  endif
          # endif
          #endif

Comment 9

[Alex Fritze]

Attachment: configure.in patch

Roland, 
Try if this works.

Comment 10

[Roland Mainz]

(In reply to comment #9)
> Roland, 
> Try if this works.

I don't have autoconf installed on Solaris but it looks that the patch should
catch the issue...

Comment 11

[Mike Kaply [:mkaply]]

Comment on attachment 144027 [details] [diff] [review]
configure.in patch

a=mkaply

Comment 12

[Alex Fritze]

checked in