Author belopolsky
Recipients belopolsky, doko, georg.brandl, haypo, loewis, mark.dickinson, pitrou, r.david.murray, rhettinger, rnk, ruseel, tim.golden, tim_one, tlesher
Date 2010-12-13.19:29:39
SpamBayes Score 4.33653e-13
Marked as misclassified No
Message-id <1292268581.13.0.183673441254.issue9012@psf.upfronthosting.co.za>
In-reply-to
Content
> it sounds like the patch is ready for commit.

there are two pending patches here:

time.diff - a hack that makes _time.c look like a module while it is not.

add_time_to_vc8_build.diff - a patch for VC 8.0 project file.

I don't like time.diff because it fixes the symptom rather than the core problem.  I believe the code that is shared between C modules should be segregated into a true module which would expose its C API via capsule mechanism.  Others suggested that this is an overkill and that capsule mechanism is not suitable for the stdlib.  On the latter argument, I would like to point that unicodedata already uses capsule mechanism to make itself available to Python's builtin str type.

As I mentioned in msg109219, this situation is not unique to time/datetime.  The math and cmath modules similarly use linker tricks to share code in _math.c.

Unassigning because I don't have an informed opinion on add_time_to_vc8_build.diff and cannot move the _time.c issue forward until similar problem is solved for _math.c which I used as the basis for _time.c design.
History
Date User Action Args
2010-12-13 19:29:41belopolskysetrecipients: + belopolsky, tim_one, loewis, georg.brandl, rhettinger, doko, mark.dickinson, pitrou, haypo, ruseel, tim.golden, tlesher, r.david.murray, rnk
2010-12-13 19:29:41belopolskysetmessageid: <1292268581.13.0.183673441254.issue9012@psf.upfronthosting.co.za>
2010-12-13 19:29:39belopolskylinkissue9012 messages
2010-12-13 19:29:39belopolskycreate