I guess it would be helpful if you could provide the profiles that caused the
crash. I've coding a higher level OSX LCMS wrapper for 2 weeks now and did not
have a crash.
Mark
On 10.06.2010, at 22:32, Paul Cezanne wrote:
> I'm in the middle of porting my Little CMS application to 2.0 and once I got
> it to compile I've encountered a crash. This is under OS X, running 10.6.2,
> with Xcode 3.2.1 as the build environment. I'm running from sources found in
> lcms2-2.0a.tar.gz.
>
> This is what my stack looks like:
>
> #0 0x00049bff in cmsPipelineCheckAndRetreiveStages at cmslut.c:131
> // line #s are not exact since I had put some debugging code in,
> // I've removed the debugging code below and indicated the line
> #1 0x00030fd5 in FixWhiteMisalignment at cmsopt.c:475
> #2 0x00032622 in OptimizeByResampling at cmsopt.c:661
> #3 0x0003278c in _cmsOptimizePipeline at cmsopt.c:1620
> #4 0x000467be in cmsCreateExtendedTransform at cmsxform.c:541
> #5 0x00046b7e in cmsCreateMultiprofileTransformTHR at cmsxform.c:643
> #6 0x00046bf6 in cmsCreateTransformTHR at cmsxform.c:684
> #7 0x00046c43 in cmsCreateTransform at cmsxform.c:694
> #8 0x00016b59 in MakeColorTransform at FineEyeCMS.cpp:341
>
> You can see the crashing line here:
>
> cmsBool CMSEXPORT cmsPipelineCheckAndRetreiveStages(const cmsPipeline* Lut,
> cmsUInt32Number n, ...)
> {
> va_list args;
> cmsUInt32Number i;
> cmsStage* mpe;
> cmsStageSignature Type;
> void** ElemPtr;
>
> // Make sure same number of elements
> if (cmsPipelineStageCount(Lut) != n) return FALSE;
>
> va_start(args, n);
>
> // Iterate across asked types
> mpe = Lut ->Elements;
> for (i=0; i < n; i++) {
> // Get asked type
> Type = va_arg(args, cmsStageSignature); // <<------ crashes here
> if (mpe ->Type != Type) {
>
>
>
>
> It certainly looks like the call to va_arg() is what is dying. I do have a
> complier warning, cmsStageSignature is promoted to 'int' when passed through
> '...'
>
> That may just be a red herring or it may be important.
>
> Here is the caller:
>
>
>
> // Locate the node for the white point and fix it to pure white in order to
> avoid scum dot.
> static
> cmsBool FixWhiteMisalignment(cmsPipeline* Lut, cmsColorSpaceSignature
> EntryColorSpace, cmsColorSpaceSignature ExitColorSpace)
> {
> cmsUInt16Number *WhitePointIn, *WhitePointOut;
> cmsUInt16Number WhiteIn[cmsMAXCHANNELS], WhiteOut[cmsMAXCHANNELS],
> ObtainedOut[cmsMAXCHANNELS];
> cmsUInt32Number i, nOuts, nIns;
> cmsStage *PreLin = NULL, *CLUT = NULL, *PostLin = NULL;
>
> if (!_cmsEndPointsBySpace(EntryColorSpace,
> &WhitePointIn, NULL, &nIns)) return FALSE;
>
> if (!_cmsEndPointsBySpace(ExitColorSpace,
> &WhitePointOut, NULL, &nOuts)) return FALSE;
>
> // It needs to be fixed?
>
> cmsPipelineEval16(WhitePointIn, ObtainedOut, Lut);
>
> if (WhitesAreEqual(nOuts, WhitePointOut, ObtainedOut)) return TRUE; //
> whites already match
>
> // Check if the LUT comes as Prelin, CLUT or Postlin. We allow all
> combinations
> if (!cmsPipelineCheckAndRetreiveStages(Lut, 3, cmsSigCurveSetElemType,
> cmsSigCLutElemType, cmsSigCurveSetElemType, &PreLin, &CLUT, &PostLin))
> if (!cmsPipelineCheckAndRetreiveStages(Lut, 2,
> cmsSigCurveSetElemType, cmsSigCLutElemType, &PreLin, &CLUT))
> if (!cmsPipelineCheckAndRetreiveStages(Lut, 2,
> cmsSigCLutElemType, cmsSigCurveSetElemType, &CLUT, &PostLin))
> if (!cmsPipelineCheckAndRetreiveStages(Lut, 1,
> cmsSigCLutElemType, &CLUT))
> return FALSE;
>
>
>
> Lut seem to be ok, CLUT is still NULL.
>
> I had to talk my boss into the 2.0 port and he really doesn't want me
> debugging this so unless it is quick I'm going to have to put this aside for
> now. But I know I'll need to look at again someday. Thank you and merci!
>
> Paul
>
>
>
>
>
> ------------------------------------------------------------------------------
> ThinkGeek and WIRED's GeekDad team up for the Ultimate
> GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
> lucky parental unit. See the prize list and enter to win:
> http://p.sf.net/sfu/thinkgeek-promo_______________________________________________
> Lcms-user mailing list
> Lcms-user@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/lcms-user
------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit. See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo
_______________________________________________
Lcms-user mailing list
Lcms-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/lcms-user
