Show quoted text>>>>> On Mon, 06 Aug 2007 11:41:07 -0400, "Mark Nodine via RT" <bug-Slay-Maker@rt.cpan.org> said:
> I sent an email to the original developer of Slay::Maker (I only updated
> it to get rid of pseudohashes, which are no longer supported in perl
> 5.9.5 and to make a few minor enhancements). Here is his response:
Show quoted text > Hmm, comments inline below, in case they help.
>>
>> t/make....# Test 45 got: "1" (t/make.t at line 372)
>> # Expected: ""
>> # t/make.t line 372 is: ok( join(',',@new_stats[8,9]) eq
>> join(',',@file_0_stats[8,9]), '' ) ;
> This is testing to see if file 0 has been re-made; it should not have
> been, but has been.
This statement seems to contradict the fact that the test immediately
before this test has a C<< force=>1 >>.
Show quoted text > If you can reproduce this, try removing the "eq"
> and replacing it with a "," in those two tests so it'll dump out the
> numbers on a good test and a bad test, that way we can be sure it's the
> mod_time and not the size that is changing.
This I did and this is the result:
% make test [CPANSHELL]
make[1]: Entering directory `/home/k/.cpan/build/Slay-Maker-0.05-BZgZMp'
PERL_DL_NONLAZY=1 /home/src/perl/repoperls/installed-perls/perl/pSYX0BQ/perl-5.8.0@31843/bin/perl "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t
t/make....NOK 45/60# Test 45 got: "1189568219,1189568220" (t/make.t at line 372)
# Expected: "1189568219,1189568219"
# t/make.t line 372 is: ok( join(',',@new_stats[8,9]), join(',',@file_0_stats[8,9]) ) ;
t/make....NOK 48/60# Test 48 got: "16;1189568219;1189568220" (t/make.t at line 388)
# Expected: "16;1189568219;1189568219"
# t/make.t line 388 is: ok( join(';',@new_stats[7..9]), join(';',@file_0_stats[7..9]) );
t/make....FAILED tests 45, 48
Failed 2/60 tests, 96.67% okay
Failed Test Stat Wstat Total Fail List of Failed
-------------------------------------------------------------------------------
t/make.t 60 2 45 48
Failed 1/1 test scripts. 2/60 subtests failed.
Files=1, Tests=60, 2 wallclock secs ( 0.58 cusr + 0.18 csys = 0.76 CPU)
Failed 1/1 test programs. 2/60 subtests failed.
make[1]: *** [test_dynamic] Fehler 255
make[1]: Leaving directory `/home/k/.cpan/build/Slay-Maker-0.05-BZgZMp'
So it is not the size but the mtime.
Show quoted text >> # Test 48 got: "1" (t/make.t at line 388)
>> # Expected: ""
>> # t/make.t line 388 is: ok( join(';',@new_stats[7..9]) eq
>> join(';',@file_0_stats[7..9]), '' );
> Same here.
> It's very likely that the test script is, 20% of the time, running
> either so fast or so slow that it's updating files "within the same
> second" sometimes and not other times and Slay::Maker is thereby
> sometimes updating a file and sometimes not. I can't recall which
> assumption Slay::Maker makes if a master file A0 and a dependent file A1
> (where A1 is built from A0) have identical filesystem dates. One could
> assume that A1 is up to date and need not be built, or one could assume
> (conservatively) that it it not up to date and rebuild it just to be
> sure. While the latter is rare, it comes in to play when you can run
> multiple make commands within the same second and updating A0 in
> between. And to think that some older filesystems have a *2* second
> resolution on their mod_time!
Show quoted text > A well-placed sleep in make.t could make this problem become stable (by
> preventing it or by making it always occur).
That would be after line 365. I added the sleep 1 there:
362 ## See if utime and atime are not restored when the size changes.
363 sub {
364 tweak_file_0() ;
365 age_file_0() ;
366 sleep 1;
367 $m->make( $file_0_name, {force=>1, detect_no_size_change=>1} ) ;
368 ok( $m->output, "created $file_0_name" ) ;
369 },
With this change the test always fails. I suggest removing the two
tests, they seem plain wrong.
--
andreas
