On Sun, May 01, 2016 at 09:37:50PM +1200, Michael Cree wrote: > subversion FTBFS on Alpha due to a test suite failure [1]:
Yeah, I had noticed the failure but didn't have access to an alpha to investigate. > START: utf-test > PASS: lt-utf-test 1: test is_valid/last_valid > PASS: lt-utf-test 2: test last_valid/last_valid2 > PASS: lt-utf-test 3: test svn_utf_cstring_to_utf8_ex2 > PASS: lt-utf-test 4: test svn_utf_cstring_from_utf8_ex2 > PASS: lt-utf-test 5: test svn_utf__normcmp > PASS: lt-utf-test 6: test svn_utf__glob > PASS: lt-utf-test 7: test svn_utf__fuzzy_escape > PASS: lt-utf-test 8: test svn_utf__is_normalized > svn_tests: E235000: In file > '/«PKGBUILDDIR»/subversion/tests/libsvn_subr/utf-test.c' line 807: assertion > failed (0 == strcmp(result->data, tc->result)) > FAIL: lt-utf-test 9: test svn_utf__utf{16,32}_to_utf8 > END: utf-test > ELAPSED: utf-test 0:00:00.603655 > > This failure occurs because the test suite allocates char aligned > strings in function test_utf_conversions() of file > subversion/tests/libsvn_subr/utf-test.c and passes those strings > to functions svn_utf__utf16_to_utf8() and svn_utf__utf32_to_utf8() > which are defined (in file subversion/libsvn_subr/utf.c) to take > apr_uint32_t * and apr_uint16_t * args respectively. Thanks for the analysis and working on a patch. I'll note that _Alignas, if it had worked, would likely have been too new a compiler feature to use. It looks like Debian's ia64 porterbox is still running, so I can try reproducing the problem there (using prctl to force SIGBUS) and then fiddling around with the code. Cheers, -- James GPG Key: 4096R/331BA3DB 2011-12-05 James McCoy <james...@debian.org>