Bugs in weekly archiving - mailman 1.1?
I have a problem that has shown up since I changed to mailman 1.1 (from 1.0). Coincidently the platform the system ran on was upgraded to RH 6.1 linux (this was all due to a hardware failure) - I think this upgraded python and glibc.
Heavy traffic lists I archive on a weekly basis. At the start of each month it produces invalid datestamps, ie today Week-of-Mon-20000200 Week-of-Mon-200001-4 Week-of-Mon-199912-1
In the source, Mailman/Archiver/HyperArch.py there is the code:-
elif self.ARCHIVE_PERIOD == 'week': datetuple = list(datetuple) datetuple = datetuple - datetuple # subtract week
day # # even if the the day of the month counter is negative, # we still get the right thing from strftime! -scott # return time.strftime("Week-of-Mon-%Y%m%d", tuple(datetuple))
I've test bedded this code under perl (which I can write rather better) and the system/library strftime() behaves as "scott" says - so putting in the -4th of Jan 2000 gives strftime output of 27th Dec 1999.
I'm starting to wonder if python put a strftime implementation into its library functions which does not match these assumptions... although it does appear to link against one in the C libraries. All in all I'm a tad confused over what is causing this problem.
I could work round this by changing that code to read datetuple=time.gmtime(date - (datetuple * 86400))
which will always work - although its more costly on CPU. Its also working around a problem that should not be there :-)