remove absolute filepath generation from ../lib
it's actually no better than plain '../lib', and will probably gain an
AthensMark or two
WN
Isn't this the same as
Item1177 (or at least the work you and Babar were doing on setlib.cfg?)
--
CrawfordCurrie - 08 Mar 2009
I have no idea what an
AthensMark is (might be some left-overs from the Athens release, but hey, I wasn't born when Athens was released, was I?), but the idea there is that ../lib is relative to the current working directory, which is very unlikely to be accurate (think some external tools, some mod_perl or FastCGI, etc...).
So what I did for
Item1177 was try to guess the full path from the filename, and use fullpath/../lib, which in 99% of the cases should expand to
/var/to/foswiki/bin/../lib
, which should be right. In some few cases, it can expand to
../lib
, which was the cause of the last bug I fixed yesterday (as I had an extra /, it expanded to
/../lib
), and this happens when everything is done relatively (like
perl view ...
).
So I really don't see the problem with the fix I did (fix was because extender.pl calls installation from the root directory, and it's the same with all installers, therefore it should use
lib
and not
../lib
.
Will, please elaborate, I really don't get your point.
Also, how come it was reported by TWiki::Main.WillNorris?
--
OlivierRaginel - 08 Mar 2009
it was reported by TWiki::Main.WillNorris because i reported this bug
over 3 years ago (Sun, 26 Feb 2006)---i wrote that line of code originally and wanted to improve it. i was hoping to get rid of the
Cwd
call for speed reasons, because i had been led to believe this call was slow (when we did profiling for dakar).
so, let's call this a "duplicate" of
Item1177 and i'll close this ancient wishlist item.
--
WillNorris - 09 Mar 2009
and
pop, it's back open again.
[02:22] * Babar agrees I really don't like the "use" Cwd there
[02:22] <Babar> it should be a require
--
WillNorris - 09 Mar 2009