Re: 'lintpkgsrc' comparing checksums on like-named patches in different packages

[David Sainty <david.sainty%gmail.com@localhost>]

On 14/09/15 22:43, Greg Troxel wrote:
> "John D. Baker" <jdbaker%mylinuxisp.com@localhost> writes:
>
> > With the freeze for pkgsrc-2015Q3 imminent, I'm going ahead and starting
> > the update process now.  After updating to HEAD, I typically run
> > 'lintpkgsrc' to clean out the old and fetch the new distfiles and such.
> >
> > Something I've noticed during these runs is complaints about checksum
> > mismatches on package patch files where the comparison is between two
> > distinctly different packages:
> >
> > checksum mismatch between 'SHA1' for 'patch-libstdc++-v3_config_os_bsd_netbsd_ctype__inline.h' in lang/gcc44 and lang/gcc34
> > checksum mismatch between 'SHA1' for 'patch-libstdc++-v3_config_os_bsd_netbsd_ctype__noninline.h' in lang/gcc44 and lang/gcc34
> I have been seeing this too.

A quick look at the code suggests that lintpkgsrc is helpfully checking 
that patches with exactly the same (long style) name but in the distinfo 
files of different packages have exactly the same hash.  I guess it's 
for packages that set PATCHDIR but maintain their own distinfo, like 
gmplayer, which refers to mplayer patches.

In this case it looks like a harmless warning.  Actually, the 
mismatching patch files are semantically identical in these two specific 
cases, so a cheap workaround would be to make them byte-wise identical too.

Or fix lintpkgsrc :)