[lsb-discuss] Linux Application Checker: FHS violation and RPMTAG_PROVIDENAME count mismatch

Mats Wichmann mats at wichmann.us
Tue Jan 28 00:38:50 UTC 2014


On 01/27/2014 12:59 AM, Andrey Makhalkin wrote:
> Denis,
> 
> Test does not pass.
> 
> If I specify single Provides, rpmbuild generates package with 3 provides:
> Samsung-spl
> Samsung-spl
> Samsung-spl(x86-64)
> 
> With no Provides: in spec and AutoProv: no there are 2 provides:
> Samsung-spl
> Samsung-spl(x86-64)


the limitation to only one provide in LSB dates back a very long way,
when it was assumed you would pick a package name, register it, and to
keep the namespace clean, you would not provide any other names.

when there's a shared library in the package it looks like modern
rpmbuild's can't easily be restricted from providing more than one name
(you'd have to go fiddle with the scripts it calls). I've at least seen
the same on Red Hat / Fedora branches, haven't really noticed if other
variants are doing the same.

informal discussion suggests maybe the old restriction doesn't make that
much sense any longer.  someone will file a bugzilla entry to prompt a
bit more discussion about this.  for the moment, lsbpkgchk is trying to
enforce what the spec currently says.


-- mats


More information about the lsb-discuss mailing list