On Tuesday, 15 September, 2015 10:41 CEST, Gordan Bobic <[email protected]> 
wrote: 
 
> On 2015-09-15 09:09, Jacco Ligthart wrote:
> > On 14-09-15 15:41, Gordan Bobic wrote:
> >> Here is the patch I wrote to get it to compile:
> >> http://ftp.redsleeve.org/pub/patches/tmp/gcc-rtti-arm.patch
> >> 
> >> You will find the modified spec file in the same directory, that also
> >> needed changing (no idea what the deal is with:
> >> %global _gnu 7E
> >> happens to be, it makes no sense to me, and it was breaking the build
> >> badly.
> > 
> > OK, I got gcc-4.8.2 to compile.
> > I installed it in a mock chroot and "fixed" FF to not use the bundeled 
> > gcc.
> 
> You mean you built the bundled gcc, installed it in the chroot, and are
> building FF in that chroot so you don't have to rebuild gcc on every> 
> attempt? I found that doing that gave me a different FTBFS than using> the 
> (fixed) (re)bundled version.

That's exactly what I meant. 
I did not try the bundled method, so I don't know if there is a difference (for 
FF)

> 
> > the FF build now fails on me with these errors:
> > 
> > <error_msg>
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/BaselineDebugModeOSR.cpp:
> > In function 'void
> > EmitBaselineDebugModeOSRHandlerTail(js::jit::MacroAssembler&,
> > js::jit::Register, bool)':
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/BaselineDebugModeOSR.cpp:1080:23:
> > error: reference to 'R0' is ambiguous
> >          jumpRegs.take(R0);
> >                        ^
> > /usr/include/sys/ucontext.h:43:3: note: candidates are: <anonymous > > 
> > enum> R0
> >    R0 = 0,
> >    ^
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/arm/BaselineRegisters-arm.h:26:39:
> > note:                 constexpr const js::jit::ValueOperand js::jit::R0
> >  static MOZ_CONSTEXPR_VAR ValueOperand R0(r3, r2);
> >                                        ^
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/BaselineDebugModeOSR.cpp:1081:23:
> > error: reference to 'R1' is ambiguous
> >          jumpRegs.take(R1);
> >                        ^
> > /usr/include/sys/ucontext.h:45:3: note: candidates are: <anonymous > > 
> > enum> R1
> >    R1 = 1,
> >    ^
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/arm/BaselineRegisters-arm.h:27:39:
> > note:                 constexpr const js::jit::ValueOperand js::jit::R1
> >  static MOZ_CONSTEXPR_VAR ValueOperand R1(r5, r4);
> >                                        ^
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/BaselineDebugModeOSR.cpp:1091:23:
> > error: reference to 'R1' is ambiguous
> >          masm.popValue(R1);
> >                        ^
> > /usr/include/sys/ucontext.h:45:3: note: candidates are: <anonymous > > 
> > enum> R1
> >    R1 = 1,
> >    ^
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/arm/BaselineRegisters-arm.h:27:39:
> > note:                 constexpr const js::jit::ValueOperand js::jit::R1
> >  static MOZ_CONSTEXPR_VAR ValueOperand R1(r5, r4);
> >                                        ^
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/BaselineDebugModeOSR.cpp:1092:23:
> > error: reference to 'R0' is ambiguous
> >          masm.popValue(R0);
> >                        ^
> > /usr/include/sys/ucontext.h:43:3: note: candidates are: <anonymous > > 
> > enum> R0
> >    R0 = 0,
> >    ^
> > /builddir/build/BUILD/firefox-38.2.1/mozilla-esr38/js/src/jit/arm/BaselineRegisters-arm.h:26:39:
> > note:                 constexpr const js::jit::ValueOperand js::jit::R0
> >  static MOZ_CONSTEXPR_VAR ValueOperand R0(r3, r2);
> >                                        ^
> > make[5]: *** [Unified_cpp_js_src3.o] Error 1
> > </error_msg>
> 
> Just out of interest, did you use my modified FF spec file?
Nope, I think you refer to FF version 24. I thought that the version difference 
was so big that it made sense to start from scratch (and because the same 
builds just fine on 7.1)

> And did you apply the armv5tel patch for the configure script?
same, no

> 
> Looking at the errors here, they don't look familiar, I think my FTBFS 
> was
> different. I have a sneaky suspicion it might be related to some headers
> somewhere (/usr/include/sys/ucontext.h ?)
You could be right, ucontext.h defines R0 and R1 etc, where the same things are 
called REG_R0 etc on RHEL6 (and RSEL7)

> As a possible alternative bodge, is there a configure option to disable
> jit support completely at build time?
> 
> > I wonder how this can fail while the same code compiles on RSEL7 :(> > 
> > google does not hold any answers today.
> > Probably the best course of action is to review all patches included in
> > FF for 6.7 vs those in 7.1 and see if something points in the direction
> > of these errors.
> 
> I looked into that, but I couldn't spot anything too obvious. The spec 
> file
> was very similar and I don't recall any difference in the patches. But 
> it's
> worth re-checking with a fresh pair of eyes.
> 
> > but first I'm going to test TB.
> 
> So TB built OK?
Nah, I meant test if it builds. test done, it doesn't. It fails similar to FF 
both when I bundle gcc and when I don't.

So apparently, first step is to verify if ucontext.h is "correct". is it 
similar to older glibc-header files? etc.

Jacco
 


_______________________________________________
users mailing list
[email protected]
http://lists.redsleeve.org/mailman/listinfo/users

Reply via email to