Re: glibmm build issues with gcc 4.8

[Havard Eidnes <he%NetBSD.org@localhost>]

> 
> coypu%sdf.org@localhost writes:
> 
>> you seem to have flawed perceptions about the state of pkgsrc GCC, which
>> until recently was barely functional for netbsd. the platforms that
>> can't build pkgsrc GCC aren't esoteric, they include, e.g. netbsd/arm.
> 
> This is the first time I've seen somebody say that so clearly.   Perhaps
> you can send a brief summary of what works and what doesn't.

netbsd-6/macppc:

ftp://malus.urc.uninett.no/reports/2017Q2/20170914.0941/gcc5-5.4.0nb3/build.log

checking for double-to-integer conversion bug... yes or failed to exec (exit status is 191)
configure: error: double-to-integer conversion is incorrect.
You need to use another compiler (or lower the optimization level).
Makefile:5781: recipe for target 'configure-stage1-mpfr' failed
gmake[2]: *** [configure-stage1-mpfr] Error 1

netbsd-7/macppc:

ftp://granny-smith.urc.uninett.no/reports/2017Q3/20171110.0844/gcc5-5.4.0nb6/build.log

checking for C compiler default output file name... 
configure: error: in `/usr/pkgsrc/lang/gcc5/work/build/powerpc--netbsd/soft-float/libgomp':
configure: error: C compiler cannot create executables
See `config.log' for more details.
Makefile:20489: recipe for target 'configure-stage1-target-libgomp' failed
gmake[2]: *** [configure-stage1-target-libgomp] Error 1

NetBSD/macppc 8.0 builds gcc 5 to completion:

morden% pwd
/ftp/pub/pkgsrc/packages/NetBSD/powerpc
morden% ls -l */All/gcc5-*
-rw-r--r--  1 he  netbsd  89635440 Nov 13 09:48 8.0_2017Q3/All/gcc5-5.4.0nb6.tgz
-rw-r--r--  1 he  netbsd  10226381 Dec  5 20:09 8.0_2017Q3/All/gcc5-libs-5.4.0nb6.tgz
morden% 

It's entirely possible that the root cause for the -6 and -7
build failures are in the base operating system, e.g. related to
float handling...

>> I've been fighting to make this work for forever and everyone just seems
>> to think this works, how do you think it would magically happen? nobody
>> in netbsd seems to care about upstreaming GCC work and leaving it up to
>> anyone who wants functional packages.
>
> This is a good point, but making pkgsrc gcc work is largely separable
> from whether systems with old base should use newer pkgsrc gcc when it
> does work.

Well, it's a matter of whether we essentially disable support
for lots of platforms by insisting on the use of a modern GCC (if
insistence is what's being suggested).

Regards,

- Håvard