wdiff-bugs
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [wdiff-bugs] wdiff-1.0.2 test 3 (pager) failures


From: Martin von Gagern
Subject: Re: [wdiff-bugs] wdiff-1.0.2 test 3 (pager) failures
Date: Tue, 15 Nov 2011 15:06:12 +0100
User-agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; de; rv:1.9.2.13) Gecko/20101207 Lightning/1.0b2 Thunderbird/3.1.7

Dear Nelson,

there are known bugs in screen. The sed script used to work around those
bugs isn't fully portable. That incompatibility is addressed by
http://bzr.savannah.gnu.org/lh/wdiff/trunk/revision/164 . The next wdiff
release will contain that fix.

If possible, please try applying that patch to the systems causing you
trouble. In case some of them aren't addressed by that patch, I'd like
to find a fix and include it in the next release.

Am 11.11.11 19:36, schrieb Nelson H. F. Beebe:
> On Solaris 11, and several other systems, [...]
> Examination of the tests/testsuite.log file showed that the
> lines differed: one had CR LF terminators, and the other LF
> terminators.

That's what the sed script was supposed to fix. In cases where it
didn't, it usually removed occurrences of the letter "r", but the above
changeset should take care of this as well. I assume that the screen log
hat CR LF, and the expected output had LF only. If it's the other way
round, I'd have to figure out why the test suite shell script writes CR
LF on those systems, or perhaps simply remove those afterwards.

> I switched from tcsh to ksh, and reran the tests, and the
> same failure occurred.  Then I switched to bash, and the
> tests passed.

Meaning you ran the test suite using bash? Strange, and an ill omen: it
would appear that I cannot rely on line terminators for documents
included in the test suite, which would make this whole autotest
approach seem questionable at least.

>  No recompilation was done in those experiments.
> I therefore concluded that the failure is some minor shell
> issue, and installed wdiff-1.0.2 on that system.

Should be safe enough. The screen test is fairly new, but mostly testing
old and well established behaviour. Unfortunately I can't tink of an
easy and portable way to test "istty" behaviour without it.

> On Mac OS X 10.5.8 PowerPC and x86_64, the same test fails, but for a
> different reason, and switching shells does not get a successful test:
> 
>       +env: wdiff: No such file or directory

OK, I can reproduce this. Seems that OS X screen does not inherit its
PATH from the parent process. Might be some configuration problem, or
something else. Including the PATH in the env command line arguments
works around this one. Will commit a fix shortly.

> On MirBSD, NetBSD, OpenBSD, and Fedora 15 (all x86), the third test
> fails with this error:
> 
>       testsuite: error: cannot find screen
> 
> I ran "yum install screen" on the Fedora system, and then got
> wdiff-1.0.2 tests to pass.
> 
> I have not yet been successful in building screen on the three BSD
> flavors.
> 
> Can the tests be done without having the screen utility available?

I had intended the test to be simply skipped if screen wasn't present.
But apparently this got broken at some point.

Judging from a quick experiment, this appears to be due to the AT_TESTED
invocation, which is intended to include additional data on screen in
the build log. Apparently, that macro causes some action out of line, so
that even if I skip the whole screen-related test, the whole suite will
still fail. Will include a fix for this as well in the next release.

Thanks for your contribution,
 Martin

Attachment: signature.asc
Description: OpenPGP digital signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]