On 23/08/10 16:39, Gaetan Bisson wrote:
[2010-08-23 16:46:09 +0200] Pierre Schmitz:
On Mon, 23 Aug 2010 11:21:20 +0200, Gaetan Bisson <bisson@archlinux.org> wrote:
This is to ensure $HOME is a writable directory during makechrootpkg runs. I suspect this is helpful to build various packages - in the example I have in mind (uim in [extra]), the home directory was determined by getpwuid(getuid())->pw_dir, so simply exporting a new HOME environment variable in the PKGBUILD wasn't enough.
I would say if a build or package function tries to write file in a user's home directory it should be considered broken and the PKGBUILD should be fixed.
I completely agree that it's broken to write to $HOME at build time.
Nevertheless, when a package does that and uses getpwuid() to determine the home directory, there is no way that I know of to deal with this solely in the PKGBUILD: the source must be patched, for instance to use getent("HOME") rather than getpwuid(), which is, in my opinion, not as satisfying a fix as simply setting ~nobody to /build.
That said, well, both fixes work... :)
Is this *really* bad, in fact so bad that upstream should be told about it, since it doesn't only affect packagers? (AFAICS the build system provided by upstream only works for individual users installing a personal build.) Has this already been communicated to upstream? /M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe