[aur-general] Advice needed maintaining boost-compute

Eli Schwartz eschwartz93 at gmail.com
Fri Nov 11 16:51:58 UTC 2016

On 11/11/2016 11:37 AM, Julien JPK wrote:
> Hello!
> I recently started maintaining small AUR packages (basically just
> updating version numbers on git packages and so on), and I came into a
> little situation when updating the boost-compute [1] package.

Do you mean, pushing a new pkgver after releases? This is only a user
convenience, no need to hunt down packages to do this (and definitely
please don't nudge maintainers to do so).

> The package is basically a header-only library, typically providing
> files under /usr/include/boost/compute. The package can be installed
> without trouble using the usual makepkg + pacman -U combo.

No it can't, since the package depends on boost and conflicts with files
in boost. It has nothing to do with yaourt, yaourt runs makepkg under
the hood to build the package, and runs pacman under the hood to install
the package.

pacman will not install this package unless you use the "--force" or
"--nodeps" flags, which is a sign that something is deeply wrong.

> However when I tried to install the package using yaourt, I realised a
> little something: the package depends on boost [2], which now seems to
> provide the boost-compute library as part of its own code. This means
> that when I install boost-compute using yaourt, it first pulls boost
> from the extra repositories, which brings the include/compute/ files,
> and creates a conflict when boost-compute comes afterwards (yaourt seems
> to detect the conflict and therefore stops).
> I'm guessing the reason why the makepkg approach works is because it
> doesn't run these checks, and the boost files are simply overwritten
> (the PKGBUILD runs a simple cp). So here's what I'm hesitating about:

Thank you for validating the opinions of those who believe all yaourt
users are complete idiots, by the way...
I happen to be a yaourt user who likes to think yaourt users can be
intelligent people too, but you are hurting my efforts.

Eli Schwartz

More information about the aur-general mailing list