[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Building Subversion r4503 / self-test commit_tests 16 fails under Cygwin [Virus checked BfA]

From: <Dr.Stefan.Dalibor_at_bfa.de>
Date: 2003-02-04 17:13:57 CET

Hi,
I'm trying to install Subversion r4503 (only the local client) under
Cygwin 1.3.19-1 as CVS replacement.
Building worked fine after the following tweaks:

o Make sure that version >= 1.3.19 of the central cygwin DLL is
  installed (check by issuing `cygcheck -c|grep cygwin' from the
  shell), as cygwin-1.3.18 seems to include a non-functional entry
  point to strerror_r (configure detects the function, but each
  program trying to use it gets an access violation from the DLL
  loader).
o Deinstall the Berkeley DB libraries distributed with Cygwin (i.e.
  installed with Cygwin setup) - these are version 3.1.*, and
  Subersion refuses to use them. I tried to point Subversion's
  configure to DB4 libraries in /usr/local, but the binaries were
  linked against the system version of the DB routines (meaning almost
  all of Subversion's self-tests failed).
o Install Berkeley DB version 4.* (builds w/o problems out from the
  sources under Cygwin, but I built only static library), and build
  Subversion against the new library.

After that, Subversion passes all of it's self-tests except one:
commit_tests.py 16: commit files and dirs with URI-unsafe characters

As distributed, commit_tests.py fails before starting the test, due to
trying to generate filenames containing angle brackets. This doesn't
work under M$ OSes - the test in the Python script that checks for
the OS platform and drops the angle bracket test under M$ fails
because Python (and Subversion) are misleaded by the Cygwin Unix
emulation (see attached patch for commit_tests.py to make the platform
test functional under Cygwin).

But more serious, even after removing the angle brackets from the test
data, the test fails (see below for result details).
Now my questions are:
1. Does this mean the current stable version of Subversion is broken
   in the Cygwin environment?
2. Is a patch available from one of the development versions fixing
   the bug?
3. Or is the test failure a non-issue, provoked by the ambiguities of
   the emulation environment?

Please include me in the CC of replies, as I'm not subscribed on the
Subversion mailing list.

Thanks,
Stefan

(See attached file: subversion-commit_tests.py-patch)
(See attached file: subversion-tests.log-fail)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Received on Tue Feb 4 17:20:36 2003

This is an archived mail posted to the Subversion Dev mailing list.