[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#517650: lintian: changes made to collection scripts are not reflected unless a package changes



Raphael Geissert <atomo64+debian@gmail.com> writes:

> Right, I don't know why I didn't think about using unlink <.${coll}*>;
> instead.

I used readdir and a regex match.  The glob is probably also about as fast
with a current version of Perl that doesn't fork a shell to do the glob,
though.  I learned my aversion to using globs in Perl back when they
involved forking an external csh binary, and probably should get over it.

> Russ Allbery wrote:

>> I'm committing now with some changes along those lines.  I'm also
>> creating real files rather than symlinks to PIDs so that we can stick a
>> bit more information in the files.  It's slightly slower (three
>> syscalls instead of one), but it means that we can put the generation
>> timestamp and the Lintian version into the file, which may be useful.

> I guess the version could be of some use but the time stamp, if needed,
> can be retrieved via stat().

Yeah, it's redundant unless you do something weird like copy a lab around
without preserving timestamps.  But it seems harmless to stick it in there
for the hell of it while I'm creating a file anyway.

BTW, the parallel bits needed a fair bit of massaging to restore the
previous error handling capabilities.  (I hope I got all of that right; I
didn't create a failing collect script to test that.  I probably should
devise some way to do that in the test suite.)

-- 
Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>


Reply to: