One of the reasons is that namcap is very specific to Arch while makepkg is not. Also namcap produces quite a lot false positives.

i'm sorry but the bug you pointed out is not because namcap was not installed.

the "mistake" is there because at the time when gstreamer0.10-ffmpeg was built, gstreamer0.10-base was still using liboil and now is using orc

Hi Ionut,I didn't said it. I think the namcap will help to find mistakes on packages.

Of course the gstreamer0.10-ffmpeg don't need of namcap, but using the namcap I can see the error. It's only a example where the namcap can be very useful, OMHO.

If the makepkg run at the end, the packager will see the warning: gstreamer0.10-ffmpeg E: Dependency detected and not included (liboil) from files ['usr/lib/gstreamer-0.10/libgstpostproc.so', 'usr/lib/gstreamer-0.10/libgstffmpegscale.so'] . So, the packager can verify if it is really a error.

Ok, the pacman produces quite a lot false positives, but in a lot of cases can be very useful. And with this feedbacks, the namcap can be improved a lot.

you understand that checking with namcap was done when -ffmpeg was compiled. but because -base was recompiled after and adjusted to the new dependency(dropped liboil), now -ffmpeg is missing it?

Ok, but when -ffmpeg was compiled the liboil was necessary and it was not identified. If it now uses libo or not does not matter.

The point that I think is important is that the moment the package is being compiled, the namcap tries to find errors. It can be very useful for the packager and to improve the namcap.

The gstreamer is not the point. The point is make the work of packagers easier and reduce the number of bugs/problems.

Running namcap on every iteration of makepkg is plain stupid. Let's say I'm working on a package that passes namcap completely fine, but I want to change pkgdesc or the pkgrel for some reason. Those changes don't change anything in the functionality of the package, but makepkg would insist on running namcap another time, which consumes useless amounts of time.
Besides taking useless amounts of time, what about chroots? my chroots don't have namcap installed, and running namcap from the host system will not catch dependencies unless I install the resulting package first. Actually, makechrootpkg already runs namcap, so official packages that are built the official way already generate namcap reports when it is available.

This is a good reason for not use namcap always. You convinced me, JGC.

Another suggestion. Why not suggest the installation of namcap by the makechrootpkg if isn't available? It is used by other systems?

I do not want python and its deps in my clean chroot.

Closing as won't fix. This can be "implemented" using a simple wrapper script if really needed.