The second release candidate for the 2.4.x series was released almost a week ago with a lot more bug fixes.
Sorry, I lack motivation to deal with gnucash at the moment. Problems with 2.4.0: - Install fails if 2.2.9 is installed: http://mid.gmane.org/4A009431.3050500@buffalo.edu http://mid.gmane.org/4A2E9909.3040900@googlemail.com - Tests fail /bin/sh: line 5: 31620 Segmentation fault (core dumped) GNC_MODULE_PATH="../../../src/engine/.libs:../../../src/calculation/.libs:../../../src/app-utils/.libs:${GNC_MODULE_PATH}" GUILE_LOAD_PATH="../../../src/core-utils:../../../src/gnc-module:../../../src/engine:../../../src/scm:../../../src/app-utils:${GUILE_LOAD_PATH}" LD_LIBRARY_PATH="../../../src/app-utils/.libs:../../../src/calculation/.libs:../../../src/engine/.libs:../../../src/gnc-module/.libs:../../../src/core-utils/.libs:../../../src/libqof/qof/.libs:${LD_LIBRARY_PATH}" DYLD_LIBRARY_PATH="../../../src/app-utils/.libs:../../../src/calculation/.libs:../../../src/engine/.libs:../../../src/gnc-module/.libs:../../../src/core-utils/.libs:../../../src/libqof/qof/.libs:${DYLD_LIBRARY_PATH}" ${dir}$tst FAIL: test-load-module No gnucash being installed it reports this: WARN <gnc.engine> failed to load gncmod-backend-xml from /usr/lib64/gnucash - python detection in configure should be verified: Does python-3 really work? I could cry when it detects /usr/lib64/debug/usr/lib64: PYTHON_LDFLAGS = -L/usr/lib64/debug/usr/lib64 -lpython2.6 PYTHON_SITE_PKG = /usr/lib64/debug/usr/lib64 Does it fail if split-debug is not in FEATURES? But it installs to /usr/lib64/python2.6/site-packages/gnucash
if you attach or commit the ebuild, I can deal with python stuff, I've been doing the work for a couple of gnome packages lately so I should be able to handle it.
Gilles, the ebuild is package.mask'ed in the tree.
I've gone through extended modifications of the test suite to make it work but I can't see the end of it. All of the guile code is based on C functions that do not get affected by any environment variables making it hard to run the test suite on a system with another version of gnucash installed. Guile itself is also causing problems by reporting stupid values for linking so it needs filtering. And tests will most likely not work with FEATURES="userpriv" due to usage of $(shell ...) constructs. I'll think I'll just commit ebuild fixes to make sure it runs in most cases but there is a high probability I'll just restrict the testsuite.
Gilles, are there upstream bugs for the problems you encountered (lack of env variables etc?
Guile problem is known as bug #202205.
+ 30 Jan 2011; Gilles Dartiguelongue <eva@gentoo.org> gnucash-2.4.0.ebuild, + +files/gnucash-2.4.0-disable-dynload-test.patch, + +files/gnucash-2.4.0-fix-tests-linking.patch: + Fix multiple python slots support, fix test failures, disable python module + test due to missing files, workaround guile configuration problem, bug + #202205. Pin dependencies to appropriate slots. Add missing gnome-keyring + dependencies, sort dependencies. Re-enable parallel build since it does not + fail for me to get a bit more testing. Closes #346919 Did not unmask it yet, but feel free to do it after checking it builds fine for you as well.
(In reply to comment #7) > Did not unmask it yet, but feel free to do it after checking it builds fine for > you as well. Hi Gilles, thanks for the patches. Worked for me now (x86), although I had to set MAKEOPTS="-j1" to get it running. Regards, Torsten
@torsten, thanks for your report. Could you open a new bug report with your full build.log so I can check what's wrong with MAKEOPTS=-j>1 ?
Created attachment 262667 [details] Updated ebuild for gnucash 2.4.2 I modified the 2.4.0 ebuild to compile with gnucash 2.4.2. Modifications made: - added dependency for gwenhywfar to use "gtk" - removed linking patch - set MAKEOPTS=-j1 for installation Not tested much, works for me though.
(In reply to comment #10) > - removed linking patch did you actually run the testsuite ? > - set MAKEOPTS=-j1 for installation why ? was there any error ?
GnuCash 2.4.3 has just been released.
*** Bug 356903 has been marked as a duplicate of this bug. ***
Created attachment 264937 [details, diff] gnucash.ebuild.diff
Created attachment 264939 [details, diff] gnucash-2.4.3-fix-tests-linking.patch
Created attachment 264941 [details, diff] gnucash-2.4.3-fix-tests-linking2.patch
I tried to bump it to 2.4.3 but, during tests running, looks like their modified by /dev/null permissions from: crw-rw-rw- 1 root root 1, 3 mar 6 2011 /dev/null To simply rw for root user, leading me to a broken system :-S
(In reply to comment #17) @Pacho: Have a look here: http://forums.gentoo.org/viewtopic-t-336492-start-0-postdays-0-postorder-asc-highlight-bash+dev+null+permission+denied.html
But, what changed permissions during test phase? It's the first time I suffer it on five different machines, then, I don't think (yet) it's a "random" udev problem :-/
Reproduced again: # ls -l /dev/null -rw------- 1 root root 85 mar 11 17:48 /dev/null Is sandbox supposed to allow things change /dev permissions at src_test phase?
sandbox must allow writing to /dev/null otherwise you cant even redirect to it: foo 2>/dev/null as for differentiating between write() and chmod()/chown()/etc..., see Bug 316321 for why this isnt feasible today
also, while in the future we can perhaps catch this, it would still mean that the package would need fixing as sandbox would be configured to mark any chown/chmod attempts on /dev/null as a violation.
(In reply to comment #21) > sandbox must allow writing to /dev/null otherwise you cant even redirect to it: > foo 2>/dev/null > > as for differentiating between write() and chmod()/chown()/etc..., see Bug > 316321 for why this isnt feasible today Thanks a lot for the info :-D (In reply to comment #22) > also, while in the future we can perhaps catch this, it would still mean that > the package would need fixing as sandbox would be configured to mark any > chown/chmod attempts on /dev/null as a violation. I agree, but was a bit "surprised" about sandbox letting package to "break" /dev/null... now all is more clear to me. Regards
The /dev/null-bug ist fixed in gnucash 2.4.4 Changelog: "Bug #644556: Running the test as root happened to remove /dev/null. Check for this and also fix the respective test. Additionally, it turned out the log system never gave notice about a log output file that was unwritable. Added an appropriate CRITICAL error message. "
+*gnucash-2.4.4 (15 Mar 2011) + + 15 Mar 2011; Pacho Ramos <pacho@gentoo.org> -gnucash-2.4.0.ebuild, + -files/gnucash-2.4.0-fix-tests-linking.patch, +gnucash-2.4.4.ebuild: + Version bump with a lot of bugfixes, remove old. +
