463987 - failed to build firefox on OpenSolaris with --enable-jemalloc

Status: RESOLVED WORKSFORME

(Firefox Build System :: General, defect)

Description

[Boying Lu]

When build js.cpp when "--enable-jemalloc" is enabled, I got following errors:
/export/home/ss12/SUNWspro/bin/CC -o js  -xlibmieee -xbuiltin=%all -features=tmplife -norunpath -mt  -DDEBUG -D_DEBUG -DDEBUG_brian -DTRACING -g  js.o jsapi.o jsarena.o jsarray.o jsatom.o jsbool.o jscntxt.o jsdate.o jsdbgapi.o jsdhash.o jsdtoa.o jsemit.o jsexn.o jsfun.o jsgc.o jshash.o jsinterp.o jsinvoke.o jsiter.o jslock.o jslog2.o jslong.o jsmath.o jsnum.o jsobj.o json.o jsopcode.o jsparse.o jsprf.o jsregexp.o jsscan.o jsscope.o jsscript.o jsstr.o jsutil.o jsxdrapi.o jsxml.o prmjtime.o jstracer.o Assembler.o Fragmento.o LIR.o RegAlloc.o avmplus.o Nativei386.o jsbuiltins.o   -L./dist/lib -ljemalloc -lpthread -M /usr/lib/ld/map.noexstk -xildoff -z lazyload -z combreloc  -z ignore -R '$ORIGIN:$ORIGIN/..'          -L./dist/bin -L./dist/lib -L/export/home/brian/moz-ws/community3/builds/1110/firefox/dist/lib -lplds4 -lplc4 -lnspr4 -lpthread -ldl -lposix4 -lsocket -ldl -lm -lCrun -lCstd  -lm -lposix4 -ldl -lnsl -lsocket    
ld: fatal: library -ljemalloc: not found

Comment 1

[Boying Lu]

Attachment: patch

Comment 2

[Jeff Walden [:Waldo]]

Methinks you want jim@moz (Jim Blandy) for a build bug, leaving alone for others to confirm...

Comment 3

[Brendan Eich [:brendan]]

Comment on attachment 347248 [details] [diff] [review]
patch

Please ask Jim for build system review. Thanks,

/be

Comment 4

[Jim Blandy :jimb]

(In reply to comment #1)

Hi, Boying.  One minor question: why are you using override here?  Unless there's a specific reason you need to prevent people from setting the variable on the command line, it seems to me just assigning the variable is the right thing to do.

I think using ../.. in js/src stuff isn't right, but the existing code is certainly not right either.  Let me see how this works on Linux.

Comment 5

[Jim Blandy :jimb]

Comment on attachment 347248 [details] [diff] [review]
patch

This isn't quite the right approach, because the proposed change is only appropriate when js/src is being built as part of the larger mozilla tree.  js/src should make sense copied out on its own, and not have special knowledge of the layout of the Mozilla tree that might be around it.

I think the best approach here is for js/src/configure.in to take --with-jemalloc and --without-jemalloc switches, the latter being the default.  If specified, --with-jemalloc would optionally give the path of the directory containing the jemalloc library to link against.  If the path were omitted, then js/src should assume that -ljemalloc can be found in the normal library search path.

Comment 6

[Boying Lu]

(In reply to comment #4)
> (In reply to comment #1)
> 
> Hi, Boying.  One minor question: why are you using override here?  Unless
> there's a specific reason you need to prevent people from setting the variable
> on the command line, it seems to me just assigning the variable is the right
> thing to do.
> 
> I think using ../.. in js/src stuff isn't right, but the existing code is
> certainly not right either.  Let me see how this works on Linux.

The reason for overriding this variable is that this variable is set in rules.mk which is included near the end of Makefile.in.

Comment 7

[Ginn Chen]

This bug was gone.

We still need some changes to memory/jemalloc/Makefile.in though. See Bug 457196.
But we don't need to change js/src/Makefile.in now.