Thank you Evgeny for looking into this, and thanks for taking the time to explain the reason for the intermittent failure mode.
FWIW, the last three builds have been successful, previously it failed on approximately half of the time. Cheers, Daniel Den fre 28 feb. 2025 kl 23:15 skrev Evgeny Kotkov via dev < dev@subversion.apache.org>: > James McCoy <james...@jamessan.com> writes: > > > W: subversion/svn/commit-cmd.c:185, > > W: subversion/libsvn_client/commit.c:1096, > > W: subversion/libsvn_client/commit.c:156: > (apr_err=SVN_ERR_STREAM_SEEK_NOT_SUPPORTED) > > This error most likely happened due to an oversight in the pristine install > stream, and it should be addressed by r1924085. > > The intermittent nature of these test failures is probably due to them > triggering a less common code path where `ra_serf` needs to resend > the entire request body, requiring a stream reset. > > > W: svn: E155017: Working copy text base is corrupt > > W: subversion/libsvn_subr/checksum.c:688: > (apr_err=SVN_ERR_CHECKSUM_MISMATCH) > > W: svn: E200014: Checksum mismatch for text base of > '/home/runner/work/subversion/subversion/subversion/tests/cmdline/svn-test-work/working_copies/blame_tests-6/iota': > > W: expected: 2d18c5e57e84c5b8a5e9a6e13fa394dc > > W: actual: d41d8cd98f00b204e9800998ecf8427e > > W: > > W: subversion/libsvn_subr/checksum.c:688: > (apr_err=SVN_ERR_COMPOSED_ERROR) > > W: svn: E200042: Additional errors: > … > > W: subversion/libsvn_subr/stream.c:1476, > > W: subversion/libsvn_subr/stream.c:228, > > W: subversion/libsvn_subr/stream.c:260, > > W: subversion/libsvn_wc/wc_db_pristine.c:290: > (apr_err=SVN_ERR_STREAM_SEEK_NOT_SUPPORTED) > > This second error appears to have the same root cause, meaning the fix > should > address it as well. > > The SVN_ERR_CHECKSUM_MISMATCH error here is likely just a secondary effect > of the original issue — essentially, a misleading error added alongside > the root cause. It might be worth improving this part to avoid reporting > a checksum mismatch when we have an incomplete put. However, with > the root cause fixed, this error shouldn't happen independently. > > > Thanks, > Evgeny Kotkov >
