On Tue May 13 02:32:16 2008, ANDK wrote:
Show quoted text> Thanks for digging that out.
>
> Maybe this change alone warrants a 3.x?
This is the change that warranted 2.x.
Show quoted text> I haven't read the File::Path manpage for a while but today I discovered
> the old style / modern style stuff in it and no mention about when this
> change was implemented. I think this needs to be said somewhere and
> switching major version numbers at such points in time gives users a
> better chance to adapt their code to modern style.
Well this is the funny thing. Another of the changes that went into 2.0
(which is why I bumped the major version) was less the
modern/traditional API change and more the race prevention code that was
first patched into Debian's version, and later made it to Suse, Red-Hat
and which ever other platforms adopted Debian's patch (or was inspired
by it).
The release coincided with 5.10, and was dual-lifed at that time (or the
months running up to it).
Anyway, such smoke failures should have been occurring over the years on
those platforms. But as far as I can tell they haven't. I dunno. Maybe I
transcribed the patch incorrectly. I don't think so, but I don't have a
better explanation.
Anyway, it occurred to me today that I can restore pretty much the
original behaviour by doing the removal, and then chdir'ing to updir()
until I finally anchor the cwd in a real directory. It will mean that
the process gets its cwd changed unexpectedly, but at least it winds up
in a sane place.
Show quoted text> Thanks for doing all that, I've had my hands on File::Path many years
> ago and I recall it was quite a challenging.
Yah, it's a bit of a bugger to work with, but interesting.
Thanks,
David