I'm also now curious as to why our fat binary mpir crashes gcc. As it
is the binary, not the gmp.h file, there is a possibility that it is
actually a bug in mpir. Of course I think it is more likely a malloc
or symbol error. But we should check.

I think if we do

valgrind gcc  -c -O1 -Wall -fno-strict-aliasing -fomit-frame-pointer
 -I. -I../src/headers -I../src/graph -o
 plotport.o ../src/graph/plotport.c

we may get a hint. I don't know if building mpir with -g may help.
Probably not. In fact if we do that the problem might disappear.

Bill.

2009/11/15 Jason Moxham <[email protected]>:
>
> On Sunday 15 November 2009 02:53:57 William Stein wrote:
>> On Sat, Nov 14, 2009 at 6:48 PM, Jason Moxham <[email protected]>
> wrote:
>> > On Sunday 15 November 2009 02:24:47 William Stein wrote:
>> >> On Sat, Nov 14, 2009 at 6:17 PM, Jason Moxham
>> >> <[email protected]>
>> >
>> > wrote:
>> >> > Of course , that is the error
>> >> > LD_LIBRARY_PATH sets the path for executables so there is no need for
>> >> > it to be set when building pari (just when running it) so the only
>> >> > thing that is USING libgmp is gcc  which doesn't like it .
>> >>
>> >> This means that a temporary workaround is that in the spkg-install
>> >> script for PARI we put:
>> >>
>> >>     echo "Building and install PARI"
>> >>     TMP=LD_LIBRARY_PATH; unset LD_LIBRARY_PATH
>> >>     if [ "$UNAME" = "CYGWIN" ]; then
>> >>         # There are weird bugs in PARI's build process on Windows XP
>> >>         # under Cygwin.
>> >>         # Passing in this extra flag gets around the bug.
>> >>         $MAKE GMPINCLUDE=-I$SAGE_LOCAL/include EXTRADLLDFLAGS=-lgmp gp
>> >>     else
>> >>         $MAKE gp
>> >>     fi
>> >>     LD_LIBRARY_PATH=$TMP
>> >
>> > I think pari wont use readline if we do this
>>
>> You're right -- that rules out this approach.
>>
>
> It will detect (and use?) readline if we only clear LD_PATH for make (not make
> and configure)  so the above may work, I'll give it a go.
>
>> > , but I dont think sage uses pari
>> > like that ?
>>
>> Yes, Sage does.  It's quite common for people to install SAGE just to
>> get a PARI whose readline actually works!  They run PARI by typing:
>>
>>    sage -gp
>>
>
> I commonly use many things as hammers , but when they break that is my fault
> :)
>
>
>> > so it should be OK
>> >
>> >> Of course, I'm curious why a FAT mpir crashes GCC...
>> >
>> > Yeah , weird , just seems to be debian 4.3.2 that does it ,
>>
>> I had the same problem with half the 32-bit Linuxes I tested on.
>>
>
> Oh , right , I was under the impression that is was only on this machine.
>
>> > NOTE: gcc-4.3.2
>> > also buggers up some other packages eg gmp-4.3.x , so it may not be worth
>> > the effort trying to fix it.
>>
>> I don't understand this remark, unfortunately.
>>
>>  -- William
>>
>>
>
> >
>

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"mpir-devel" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/mpir-devel?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to