Author larry
Recipients benjamin.peterson, georg.brandl, larry, r.david.murray, serhiy.storchaka
Date 2014-01-16.00:06:57
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1389830817.33.0.648583550131.issue20274@psf.upfronthosting.co.za>
In-reply-to
Content
You're right, deprecation sounds best.  If Georg or Benjamin want the fix in earlier releases I'll split the pysqlite_connection_call into another issue, otherwise I won't bother.

As for fixing the undefined behavior in older versions: since its behavior is undefined, yes, it *could* cause a crash.  But by that token it *could* teleport you to Mars and give you a funny-looking nose.  In practice it *should* be utterly harmless, as I believe every platform supported by Python 3.4 uses the caller-pops-stack calling convention.  And we've lived with it for this long, so it doesn't seem to be hurting anything.  But like all undefined behavior I can make no guarantee.

MvL in particular comes down like a ton of bricks whenever someone proposes checking in code that's technically undefined behavior.  I've had the relevant chapter and verse of the C standard quoted at me for this exact thing (calling a function pointer using a sliiiightly different signature than the actual function).
History
Date User Action Args
2014-01-16 00:06:57larrysetrecipients: + larry, georg.brandl, benjamin.peterson, r.david.murray, serhiy.storchaka
2014-01-16 00:06:57larrysetmessageid: <1389830817.33.0.648583550131.issue20274@psf.upfronthosting.co.za>
2014-01-16 00:06:57larrylinkissue20274 messages
2014-01-16 00:06:57larrycreate