Index | Thread | Search

From:
Antoine Jacoutot <ajacoutot@bsdfrog.org>
Subject:
Re: [update] webkitgtk 2.50.0
To:
ports@openbsd.org
Date:
Fri, 26 Sep 2025 13:18:32 +0200

Download raw body.

Thread
On Wed, Sep 24, 2025 at 09:35:02PM +0100, Stuart Henderson wrote:
> On 2025/09/24 17:51, Antoine Jacoutot wrote:
> > On Wed, Sep 24, 2025 at 10:41:28AM +0100, Stuart Henderson wrote:
> > > i386 fails; errors here, full log attached.
> > > 
> > > /pobj/webkitgtk40-2.50.0/bin/c++ <<snip>>  -c /pobj/webkitgtk40-2.50.0/webkitgtk-2.50.0/Source/JavaScriptCore/llint/LowLevelInterpreter.cpp
> > > /pobj/webkitgtk40-2.50.0/webkitgtk-2.50.0/Source/JavaScriptCore/llint/LowLevelInterpreter.cpp:286:9: error: use of undeclared label 'op_instanceof_return_location'
> > >   286 |         FOR_EACH_BYTECODE_HELPER_ID(OPCODE_ENTRY)
> > >       |         ^
> > > /pobj/webkitgtk40-2.50.0/build-i386/JavaScriptCore/DerivedSources/Bytecodes.h:989:11: note: expanded from macro 'FOR_EACH_BYTECODE_HELPER_ID'
> > >   989 |     macro(op_instanceof_return_location, 0) \
> > >       |           ^
> > > /pobj/webkitgtk40-2.50.0/webkitgtk-2.50.0/Source/JavaScriptCore/llint/LowLevelInterpreter.cpp:286:37: error: use of undeclared label 'op_instanceof_return_location_wide16'
> > >   286 |         FOR_EACH_BYTECODE_HELPER_ID(OPCODE_ENTRY)
> > >       |                                     ^
> > > /pobj/webkitgtk40-2.50.0/webkitgtk-2.50.0/Source/JavaScriptCore/llint/LowLevelInterpreter.cpp:286:37: error: use of undeclared label 'op_instanceof_return_location_wide32'
> > 
> > Could you try adding the attached patch?
> > Thanks.
> 
> This gets past the above failure, but then c++ runs out of memory compiling
> ${WRKBUILD}/WebCore/DerivedSources/unified-sources/UnifiedSource-3a52ce78-40.cpp
> 
> Same if I use -O1 on that file. -O0 prints out an error talking about
> webkit being slow and says to define a macro, but then seems to go on
> to attempt to compile anyway, then again blows up OOM.
> 
> pbuild's datasize-cur is set to infinity.
> 
> So might just have to be disabled on i386.

*sigh*
That would drop lots of ports :-/
Not sure how to fix this.

-- 
Antoine