Daniel Dunbar <daniel at zuster.org> writes:> There was a bug at some point where this would happen, I've forgotten > the details but it is buried in bugzilla somewhere. It has now been > fixed, but IIRC the issue would also be resolved if you did a full > make clean. The problem was in a particular situation where every > build would essentially invalidate the table gen outputs, but not > actually update them, so the build never resolved itself.IIRC, that bug was resolved before the 2.6 fork, although what the OP says sounds like the same one.>> I am building LLVM 2.6 on VC++. I am running into this problem >> where even builds without any changes whatsoever causes rebuilding of >> certain directories like table gen of intrinsics and x86 target files. >> This then leads to build of the x86 codegen. Is this expected ? Any >> pointers to how I can avoid this?-- Óscar
Just to clarify - I can build successfully except that the following files are always determined to be out of date by VC++ intrinsics_gen - Debug Win32 X86CodeGenTable_gen - Debug Win32 LLVMOxiliCodeGen - Debug Win32 I just need to start a debug build of llc to reproduce this issue and with cmake generated files. If no one else has this problem, then it might be my environment. Otherwise I can open a bug. thanks shrey On Tue, Nov 17, 2009 at 2:23 PM, Óscar Fuentes <ofv at wanadoo.es> wrote:> Daniel Dunbar <daniel at zuster.org> writes: > >> There was a bug at some point where this would happen, I've forgotten >> the details but it is buried in bugzilla somewhere. It has now been >> fixed, but IIRC the issue would also be resolved if you did a full >> make clean. The problem was in a particular situation where every >> build would essentially invalidate the table gen outputs, but not >> actually update them, so the build never resolved itself. > > IIRC, that bug was resolved before the 2.6 fork, although what the OP > says sounds like the same one. > >>> I am building LLVM 2.6 on VC++. I am running into this problem >>> where even builds without any changes whatsoever causes rebuilding of >>> certain directories like table gen of intrinsics and x86 target files. >>> This then leads to build of the x86 codegen. Is this expected ? Any >>> pointers to how I can avoid this? > > -- > Óscar > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >
typo intrinsics_gen - Debug Win32 X86CodeGenTable_gen - Debug Win32 LLVMX86CodeGen - Debug Win32> I just need to start a debug build of llc to reproduce this issue and > with cmake generated files. > > If no one else has this problem, then it might be my environment. > Otherwise I can open a bug. > > thanks > shrey > > > On Tue, Nov 17, 2009 at 2:23 PM, Óscar Fuentes <ofv at wanadoo.es> wrote: >> Daniel Dunbar <daniel at zuster.org> writes: >> >>> There was a bug at some point where this would happen, I've forgotten >>> the details but it is buried in bugzilla somewhere. It has now been >>> fixed, but IIRC the issue would also be resolved if you did a full >>> make clean. The problem was in a particular situation where every >>> build would essentially invalidate the table gen outputs, but not >>> actually update them, so the build never resolved itself. >> >> IIRC, that bug was resolved before the 2.6 fork, although what the OP >> says sounds like the same one. >> >>>> I am building LLVM 2.6 on VC++. I am running into this problem >>>> where even builds without any changes whatsoever causes rebuilding of >>>> certain directories like table gen of intrinsics and x86 target files. >>>> This then leads to build of the x86 codegen. Is this expected ? Any >>>> pointers to how I can avoid this? >> >> -- >> Óscar >> >> _______________________________________________ >> LLVM Developers mailing list >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >> >