[Debian-med-packaging] Bug#528659: plink FTBFS: #error "Cannot determine endianness.

peter green plugwash at p10link.net
Sat May 16 12:58:17 UTC 2009


tags 528659 +patch
thanks

It seems that gcc on debian doesn't define __{BIG|LITTLE}_ENDIAN__ or 
any of the equvilents the package checks for. The package has hardcoded 
fallbacks for some architectures but not others hence the FTBFS on some 
but not all debian architectures (roughly correlated with how well known 
the architecture is).  I've added code to fallback to endian.h if all 
else fails (I included the #warning lines so that I could check the new 
code behaved correctly, you may remove them if you wish).

The patch for that issue is attatched in the file 
30_endiandetection.patch ready to be added to the quilt series. The 
patch lead to a full successfull build on armel and allowed successfull 
manual compilation of r.cpp on mips (compilation of the package as a 
whole on mips was not achived due to the other issue mentioned  below). 
In both cases my #warning lines reported the endianess I expected.

There is also a seperate issue with parse.cpp making g++ sick on 
mips(el). On the buildds it failed with assembler errors, in my qemu 
system it failed with a segfault.  Reducing the optmisation level to 02 
it still segfaulted, reducing it to 01 the build was still running after 
more time than I was prepared to leave it for (this probablly isn't 
helped by the fact that qemu-system-mips seems to crash if I give it too 
much ram)


-------------- next part --------------
A non-text attachment was scrubbed...
Name: 30_endiandetection.patch
Type: text/x-diff
Size: 1174 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/debian-med-packaging/attachments/20090516/6b57f61c/attachment.patch>


More information about the Debian-med-packaging mailing list