Ramsay Jones [off-list ref] writes:
Signed-off-by: Ramsay Jones <redacted>
---
This patch fixes the t9700 test failure on cygwin. Don't ask me to
explain why the original test.pl fails on cygwin, but passes on Linux!
My perl version:
$ perl --version
This is perl, v5.8.7 built for cygwin-thread-multi-64int
(with 1 registered patch, see perl -V for more detail)
Copyright 1987-2005, Larry Wall
Perl may be copied only under the terms of either the Artistic License or the
GNU General Public License, which may be found in the Perl 5 source kit.
Complete documentation for Perl, including FAQ lists, should be found on
this system using `man perl' or `perldoc perl'. If you have access to the
Internet, point your browser at http://www.perl.org/, the Perl Home Page.
$
[My perl version on Linux is whatever comes with Ubuntu 7.04]
Curious.
What does this give you:
$ cd t
$ sh t9700-perl-git.sh -v
with the attached patch? Unfortunately this cannot be run with "-i -v"
because 'test_external_without_stderr' uses test_external which in turn
exits upon failure when -i is given, without letting the former to report
what was given to the standard error stream (this is slightly offtopic but
pertains to debuggability of tests, so Lea is Cc'ed).
For me, cwd() is the same as Cwd->cwd (this is from Perl 5.8.8); perhaps
it behaves differently on Cygwin?
t/t9700/test.pl | 6 ++++++
1 files changed, 6 insertions(+), 0 deletions(-)
diff --git i/t/t9700/test.pl w/t/t9700/test.pl
index 851cea4..139711d 100755
--- i/t/t9700/test.pl
+++ w/t/t9700/test.pl
@@ -15,6 +15,12 @@ BEGIN { use_ok('Git') }
# set up
our $abs_repo_dir = Cwd->cwd;
+our $cwd = cwd();
+print STDERR "Cwd->cwd says '$abs_repo_dir'\n";
+print STDERR "cwd() says '$cwd'\n";
+
+exit(1);
+
ok(our $r = Git->repository(Directory => "."), "open repository");
# configJunio C Hamano wrote:
Ramsay Jones [off-list ref] writes:
quoted
Signed-off-by: Ramsay Jones <redacted>
---
This patch fixes the t9700 test failure on cygwin. Don't ask me to
explain why the original test.pl fails on cygwin, but passes on Linux!
Curious.
What does this give you:
$ cd t
$ sh t9700-perl-git.sh -v
with the attached patch?
[snip]
For me, cwd() is the same as Cwd->cwd (this is from Perl 5.8.8); perhaps
it behaves differently on Cygwin?
Yes, sorry, I should have spent a few minutes explaining the cause of the
bug and why the patch fixes it; my bad!
(I was in a bit of a hurry; I had just re-booted twice in quick succession,
once into Linux so that I could pull the patch over to test it there, before
going back to Windows, in order to format-patch and send it to the ML. Yes
it is a PITA 8-)
I haven't tried your patch because, well, I already know what it will say!
Once I determined that the problem must be in the following four lines
of the t/t9700/test.pl script:
17 our $repo_dir = "trash directory";
18 our $abs_repo_dir = Cwd->cwd;
19 die "this must be run by calling the t/t97* shell script(s)\n"
20 if basename(Cwd->cwd) ne $repo_dir;
then the problem and the solution was immediately obvious. Well, maybe that is
overstating it slightly. After all, I still indulged in a spot of "printf
debugging" just to make sure! I'm sure you can imagine what it looked like ;-)
[BTW, the reason it was obvious, is that I have met this problem *too many*
times in the past, so I'm "attuned" to it. Don't worry, I will dispense with
the rant this time :-P]
I won't repeat that here, but offer a quick script:
$ cat test-cwd.pl
use Cwd;
print "cwd(): ", cwd(), "\n";
print "Cwd::cwd(): ", Cwd::cwd(), "\n";
print "Cwd->cwd: ", Cwd->cwd, "\n";
$
on cygwin:
$ perl test-cwd.pl
cwd(): /home/ramsay
Cwd::cwd(): /home/ramsay
Usage: Cwd::cwd() at test-cwd.pl line 5.
$
and on Linux:
$ perl test-cwd.pl
cwd(): /home/ramsay
Cwd::cwd(): /home/ramsay
Cwd->cwd: /home/ramsay
$
So it is clear, the problem is caused by the "method invocation syntax".
Among other things, this mechanism auto-magically inserts a new (in this
case only) first parameter which represents the "class" Cwd to the cwd()
"method-call".
So, do a "perldoc Cwd" and take a look at the example usage. So is cwd()
a method of the Cwd class, or is it a function in the Cwd namespace?
Exactly, hence the patch. (Well, some may prefer to see the Cwd::cwd()
syntax used instead. Dunno)
So, the bug is inappropriate use of the method invocation syntax.
But that didn't explain the difference in behaviour on the two platforms,
hence my "Don't ask me to explain ..." comment.
Now it just so happens that I have the source code to perl v5.8.8 in my
home directory (no really!) on Linux, so I spent a little time rooting
around to try and explain it.
The changelog type files, perl-5.8.8/Changes and perl-5.8.8/ext/Cwd/Changes
didn't show much of interest. The module impl. in perl-5.8.8/lib/Cwd.pm was
an interesting read but didn't explain it. Then I bumped into the cygwin
specific file perl-5.8.8/cygwin/cygwin.c, part of which I include below:
---perl-5.8.8/cygwin/cygwin.c:139-157---
/* see also Cwd.pm */
static
XS(Cygwin_cwd)
{
dXSARGS;
char *cwd;
if(items != 0)
Perl_croak(aTHX_ "Usage: Cwd::cwd()");
if((cwd = getcwd(NULL, -1))) {
ST(0) = sv_2mortal(newSVpv(cwd, 0));
free(cwd);
#ifndef INCOMPLETE_TAINTS
SvTAINTED_on(ST(0));
#endif
XSRETURN(1);
}
XSRETURN_UNDEF;
}
------
So, that explains it.
ATB,
Ramsay Jones
P.S. I'm sure someone will say that "the real bug here" is that the cygwin
platform should support the method invocation syntax just like all the
other platforms. Well, I guess I don't really care that much. But If you
were to push me, I would tend to go the other way, and say the other platforms
should not allow an incorrect invocation to go unnoticed!
Whatever.