Summary: | www-apps/swish-e fails to build w/ glibc-2.8 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Diego Elio Pettenò (RETIRED) <flameeyes> |
Component: | New packages | Assignee: | Gentoo Perl team <perl> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | esigra, fauli, kentnl, pacho, sandro.bonazzola, wilscarlet |
Priority: | High | Keywords: | Inclusion |
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 225459 | ||
Attachments: |
Build log
Build log Patch for recent perls and recet ExtUtils stuff www-apps/swish-e/swish-e-2.4.4.ebuild 2.4.4-ebuild-fixed.patch |
Description
Diego Elio Pettenò (RETIRED)
2008-11-23 22:57:21 UTC
Created attachment 173087 [details]
Build log
There is a newer version 2.4.5, see bug #223589. But, it's not in portage yet and it has be released since Mon, 29 Jan 2007 19:59:04 UTC. Maybe someone should test the newer version. The version bump happened, could you please recheck if it still fails. Created attachment 184222 [details]
Build log
No still fails.
Could people with this test on their various glibc try compiling/installing/using with USE="-perl" There seems to be a bit of lobotomised code in the ebuild at present as well as broken perl code in the perl segment, and in both pasted build logs its failing in perl specific rubbish. I've tested on my box with Glib2.9 pre-releases and GCC4.4 which should really break if theres anything wrong with the C part of the deal , but its not, even swish-e-2.4.4 builds fine on it. If this is the case then we can nail down the Perl stuff separately ( or drop support if nobody cares I guess ) Created attachment 184284 [details, diff]
Patch for recent perls and recet ExtUtils stuff
This code hasn't been touched since '05. That was a long time ago. It has lots of dirty tricks in it , and it appears eval in MakeMaker, which is fine, is tripping up the build in the outer layer in ways it shouldn't.
This results in it failing the build for innane things such as
if (my $arg_max = eval { require POSIX; &POSIX::ARG_MAX }) {
$self->{_MAX_EXEC_LEN} = $arg_max;
}
else { # POSIX minimum exec size
$self->{_MAX_EXEC_LEN} = 4096;
}
Where eval is only being used for flow control, but is instead causing a fatal termination condition :/
5 Tonnes of WTF here, but it is ancient code. Also, the ebuild is full of fail too, I architectured that a bit so it at least builds in a normal order instead of doing the Perl stuff during src_install , ugh.
Hopefully this is the fix that is needed, it passes the build failures on my box. Need some more testing to confirm the modules work as expected because I'm not much of a swish-e user.
Also, this patch is for 2.4.4, because I haven't synced to get the latest, but it *should* work on 2.4.5 , that file hasn't been modified since '05 after all.
Created attachment 184286 [details]
www-apps/swish-e/swish-e-2.4.4.ebuild
swish-e-2.4.4 ebuild with a massive overhaul of stuff done to it. Please test.
It has very verbose einfo stuff in at the moment because I couldn't otherwise tell where the perly stuff started and stopped, hopefully that's not too annoying. Its got a bit of forward planning in it so its mostly ready for EAPI=2 but its still back on 0 or 1 or whatever 'undef' is.
Created attachment 184296 [details, diff]
2.4.4-ebuild-fixed.patch
For interests sake, this is the diff from the previous v of swish-e I have in my tree.
Just tested my patches and bumped ebuild to 2.4.5 , still working there for me. Doh, although, I'm a retard because I was way too tired last night. What logic is there in waiting for a bump to come through from sync when its just a filename change? srsly. Perl team, all you need is here...please find a commit monkey to get it fixed. Should be fixed in 2.4.7. Thanks. |