On Jan 13, 2012, at 11:44 AM, Ken Moffat wrote:
>> Speaking of errata, I've found a couple of packages that break when running
>> the tests in parallel:
>>
>> * gawk
>> * flex
>>
>> Executing the 'make check' with '-j 1' allows the tests to pass. I suppose
>> this would be covered in the somewhat obscure note buried in 4.5 "About
>> SBUs" where it discusses $MAKEFLAGS:
>>
> Dunno, and for the moment I don't care. I prefer to build with -j1
> so that the logs are in order *when* things break (my builds are
> development, I expect breakage, either from altering my own scripts,
> or from updates to packages). Maybe someone else knows which
> packages are like this ?
You may not care, but I'm sure other people would enjoy the time savings of
knowing which packages can be safely built and tested with -j4, and which
cannot be. I've identified the two that don't work, and everything else seems
to pass their tests when built with -j4.
(Aside from gcc & glibc, with the "known expected failures" I discussed in an
earlier post.)
Everyone wants to save time--especially when it's just spent compiling things.
I spent the time figuring it out, and I think this info is useful and could be
posted to the errata page. Even better would be a MAKEFLAGS-preserving trap
around the 'make check' statements in the book, but errata seems sufficient.
Seems like that's the place for this kind of info (e.g., coreutils time-parsing
issue breaking tests). I'm happy to do it myself, if I'm able to...
Q
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
