On MinGW64 we pass -DDLL_EXPORT -D PIC to yasm. On Cygwin64 it's -DDLL_EXPORT -DPIC.
mpn_neg_n calls mpn_com_n and mpn_add_1. The latter is a C macro. The former does not have a PIC path, but it does have SSE2 instructions. In particular is uses movdqu. Maybe there is some kind of alignment issue with malloc? Bill. On 1 April 2014 19:39, Bill Hart <goodwillh...@googlemail.com> wrote: > Here are the functions which segfault for me: > > t-dc_bdiv_q_n > t-hgcd > t-invert > t-mulmod_2expm1 > t-mulmod_2expp1 > t-neg > t-sb_bdiv_q > > I suggest we start with t-neg, as that is by far the simplest function. > There is no reason I can see for all of those to fail, unless one of the > mpn functions is broken on Cygwin64. > > The only other clue we have is this: > > 0 [sig] -bash 25560! get_proc_lock: Couldn't acquire > sync_proc_subproc for(5,1), last 7, Win32 error 0 > 60308 [sig] -bash 25560! proc_subproc: couldn't get proc lock. what 5, > val 1 > > It might be unrelated. It happened when I tried to open one of the log > files with vim immediately after. > > JP, are these the same failures for you? > > Bill. > > > On 1 April 2014 18:32, Bill Hart <goodwillh...@googlemail.com> wrote: > >> That apparently solved part of the problem, but not the segfaults with a >> shared lib. >> >> Getting closer to being able to test it myself. We'll see if it segfaults >> on my machine. >> >> Bill. >> >> >> On 1 April 2014 18:17, Jean-Pierre Flori <jpfl...@gmail.com> wrote: >> >>> >>> >>> On Tuesday, April 1, 2014 2:20:54 PM UTC+2, Bill Hart wrote: >>>> >>>> Actually, it was you that reported this. You also reported it on the >>>> Cygwin64 list (they didn't seem to know anything about the segfaults). >>>> >>>> Apparently there were problems within ld, which have presumably been >>>> resolved. >>>> >>>> On the other hand, you were able to build a shared library ok, but some >>>> (not all) of the tests segfaulted. >>>> >>>> It looks like this just never got dealt with (perhaps we dealt with the >>>> ld problem and never did anything about the other). >>>> >>>> Bill. >>>> >>> I do remember that. >>> IIRC I think the original problem was that the wrong assembly dir was >>> used (i.e. x86_64 with the AMD64 ABI instead of x86_64w with the Win64 one). >>> I don't really remember what happened next. >>> >> >> > -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to mpir-devel+unsubscr...@googlegroups.com. To post to this group, send email to mpir-devel@googlegroups.com. Visit this group at http://groups.google.com/group/mpir-devel. For more options, visit https://groups.google.com/d/optout.