Summary: | app-crypt/johntheripper-1.7.9-r6 - emake linux-x86-mmx: x86-mmx.o: In function `DES_bs_crypt': (.text+0x20): multiple definition of `DES_bs_crypt' | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Jeroen Roovers (RETIRED) <jer> |
Component: | Current packages | Assignee: | Rick Farina (Zero_Chaos) <zerochaos> |
Status: | RESOLVED OBSOLETE | ||
Severity: | normal | CC: | bircoph, crypto+disabled, david.w.noon, mbays, sam, zbox |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
app-crypt:johntheripper-1.7.9-r6:20130112-160439.log
Successful amd64 build log app-crypt:johntheripper-1.7.9-r8:20131128-020503.log |
Description
Jeroen Roovers (RETIRED)
2013-01-12 16:21:04 UTC
After additionally setting USE=sse2, it does compile properly. [ebuild R ] app-crypt/johntheripper-1.7.9-r6 USE="mmx openmp sse2 -cuda -custom-cflags -minimal -mpi -opencl" While SSE2 is fine for Pentium 4 processors, those of us with Athlon, Duron or Pentium III cannot use that option. So, MMX is still broken for us. since I am unable to replicate on amd64 can you please attach the full build log of the successful (USE=sse2) run? same USE combo works here Same USE combo, works fine here on x86. (In reply to comment #3) > since I am unable to replicate on amd64 can you please attach the full build > log of the successful (USE=sse2) run? I will take you at your word that you mean "successful" rather than "unsuccessful". I will attach the build log from my amd64 box, which worked. My problems are on an older x86 box that does not have SSE2 instructions. Created attachment 335482 [details]
Successful amd64 build log
(In reply to comment #6) > (In reply to comment #3) > > since I am unable to replicate on amd64 can you please attach the full build > > log of the successful (USE=sse2) run? > > I will take you at your word that you mean "successful" rather than > "unsuccessful". I will attach the build log from my amd64 box, which > worked. My problems are on an older x86 box that does not have SSE2 > instructions. I know it works on amd64, I can't make it fail on amd64. What I need is a log from someone on x86 who has SSE2 and can disable that flag (for a failed build log) then enable that flag (for a successful build log) and paste both so I can figure out wtf is happening here. If anyone can provide a fail and success from the same box that would be really really helpful. Otherwise I'll have to refer this issue to upstream and they will likely want the same. Whether or not DES_bs_crypt* is defined in src/x86-mmx.h depends on: #if defined(__MMX__) && defined (_OPENMP) #define DES_BS_ASM 0 The only way this gets enabled is through the ebuild, with a blanket: use openmp && OMP="-fopenmp" So it seems USE=mmx and USE=openmp should be mutually exclusive. Since the ebuild already does USE flag handling completely the wrong way, I hesitate to add a patch to work around yet another problem in the ebuild. *** Bug 492748 has been marked as a duplicate of this bug. *** Created attachment 364118 [details]
app-crypt:johntheripper-1.7.9-r8:20131128-020503.log
Here's the log of a successful compile after enabling sse2; a failed compile
without sse2 on the same box is attached to the inadvertant dup 492748.
I compiled app-crypt/johntheripper-1.8.0 successfully with USE="-openmp" CPU_FLAGS_X86="mmx sse2", |