[-- Attachment #1.1: Type: text/plain, Size: 253 bytes --] I've tagged the first release candidate for 0.31 (0.31_rc0), and uploaded signed tarballs to the usual place [1]. We need to sort out NEWS, and possibly add one or two small patches before final release of 0.31 [1]: https://notmuchmail.org/releases/ [-- Attachment #1.2: signature.asc --] [-- Type: application/pgp-signature, Size: 832 bytes --] [-- Attachment #2: Type: text/plain, Size: 0 bytes --]
[-- Attachment #1.1: Type: text/plain, Size: 349 bytes --] David Bremner <david@tethera.net> writes: > I've tagged the first release candidate for 0.31 (0.31_rc0), and > uploaded signed tarballs to the usual place [1]. > > We need to sort out NEWS, and possibly add one or two small patches > before final release of 0.31 Did the same for rc1. Mainly this fixes a silly copy-paste bug in T590-libconfig d [-- Attachment #1.2: signature.asc --] [-- Type: application/pgp-signature, Size: 832 bytes --] [-- Attachment #2: Type: text/plain, Size: 0 bytes --]
[-- Attachment #1.1: Type: text/plain, Size: 495 bytes --] David Bremner <david@tethera.net> writes: > David Bremner <david@tethera.net> writes: > >> I've tagged the first release candidate for 0.31 (0.31_rc0), and >> uploaded signed tarballs to the usual place [1]. >> >> We need to sort out NEWS, and possibly add one or two small patches >> before final release of 0.31 > > Did the same for rc1. Mainly this fixes a silly copy-paste bug in > T590-libconfig That should be T562-lib-database. Sigh. And I made the same mistake in debian/changelog. d [-- Attachment #1.2: signature.asc --] [-- Type: application/pgp-signature, Size: 832 bytes --] [-- Attachment #2: Type: text/plain, Size: 0 bytes --]
Here's a lightly edited output from git shortlog to jog your collective memories. Please send patches for NEWS for any non-trivial user visible changes. David Bremner (125): test: add known broken test for error handling on closed database lib: catch error from closed db in n_m_get_message_id lib: migrate to post Xapian 1.3.4 compact support lib: migrate from Xapian ValueRangeProcessor to RangeProcessor test: drop upgrade from v1 tests lib: add regression test for n_m_get_date; clarify API doc: replace use of environment variables with a generated config doc: add new python bindings to main documentation tree. lib: catch exceptions in n_m_get_flag, provide n_m_get_flag_st lib: add notmuch_message_has_maildir_flag_st cli/new: replace newly deprecated n_m_has_maildir_flag test: add regression test for n_m_maildir_flags_to_tags test: add broken test for n_m_remove_all_tags lib: handle xapian exception in n_m_remove_all_tags lib: fix return value for n_m_reindex lib: move deallocation of memory from n_d_close to n_d_destroy lib/n_d_get_version: catch exceptions and clarify the API lib: convert relative filenames to absolute in n_d_index_file lib: replace use of static_cast for writable databases build: upload html docs as part of release process lib: return NULL from n_d_get_default_indexopts on error lib: catch exceptions in n_directory_get_child_directories lib: catch exceptions in n_directory_get_child_files lib: fix return value for n_directory_delete Jonas Bernoulli (38): test: Deal with Emacs 27 switching to lexical scope by default test: Remove misguided emacs testing utilities test: Explicitly state that we want to sign with sender emacs: Extend face to window edge again emacs: Fix some function declarations emacs: notmuch-poll: Let the user know we are polling emacs: Improve doc-strings emacs: Autoload notmuch-jump-search only once emacs: Autoload notmuch-jump using an autoload cookie emacs: Use cl-incf where appropriate emacs: Remove notmuch-setq-local emacs: Remove notmuch-read-char-choice emacs: Drop old advices that were only need for Emacs 23 emacs: Do not abuse advice to monkey patch while testing emacs: Use new advice mechanism do advice mm-shr Jonas Witschel (1): doc: make gzipped man pages reproducible Sean Whitton (2): emacs: notmuch-search: avoid wiping out buffer-local variables emacs: Use pop-to-buffer-same-window rather than switch-to-buffer Teemu Likonen (3): Emacs: Ensure left-to-right display for message headers Emacs: Indent first header line only when indentation is turned on William Casarin (3): emacs/tree: introduce notmuch-tree-parent-buffer variable emacs/tree: enable moving to next thread in search results emacs/tree: add notmuch-tree-archive-thread-then-next Đoàn Trần Công Danh (3): configure: check for pytest with python -m pytest configure: drop check for default xapian backend T355: specify hash algorithm explicitly\r
David Bremner <david@tethera.net> writes:
> Here's a lightly edited output from git shortlog to jog your collective
> memories. Please send patches for NEWS for any non-trivial user visible
> changes.
I've sent a couple of patches [1] for things that stood out to
me. Please send me anything you think I missed ASAP; I'd like to release
the (hopefully) final release candidate tommorow.
[1]: id:20200829122310.801682-1-david@tethera.net, id:20200829122310.801682-2-david@tethera.net
[-- Attachment #1.1: Type: text/plain, Size: 364 bytes --] David Bremner <david@tethera.net> writes: > David Bremner <david@tethera.net> writes: > > I've sent a couple of patches [1] for things that stood out to > me. Please send me anything you think I missed ASAP; I'd like to release > the (hopefully) final release candidate tommorow. I've tagged 0.31~rc2, and uploaded to notmuchmail.org and debian experimental. d [-- Attachment #1.2: signature.asc --] [-- Type: application/pgp-signature, Size: 857 bytes --] [-- Attachment #2: Type: text/plain, Size: 0 bytes --]
[-- Attachment #1.1: Type: text/plain, Size: 1014 bytes --] Hi David, David Bremner <david@tethera.net> writes: > David Bremner <david@tethera.net> writes: > >> David Bremner <david@tethera.net> writes: >> >> I've sent a couple of patches [1] for things that stood out to >> me. Please send me anything you think I missed ASAP; I'd like to release >> the (hopefully) final release candidate tommorow. > > I've tagged 0.31~rc2, and uploaded to notmuchmail.org and debian > experimental. I have started seeing consistent test failures with rc2 on ppc64 and ppc64le: - on both platforms "get total messages with closed database" from T568-lib-thread fails - on ppc64 "comparing existing to exported symbols" from T360-symbol-hiding fails as well See: https://build.opensuse.org/public/build/home:dancermak:branches:devel:libraries:c_c++/openSUSE_Factory_PowerPC/ppc64/notmuch/_log and https://build.opensuse.org/public/build/home:dancermak:branches:devel:libraries:c_c++/openSUSE_Factory_PowerPC/ppc64le/notmuch/_log Any ideas what could be the issue? Cheers, Dan [-- Attachment #1.2: signature.asc --] [-- Type: application/pgp-signature, Size: 865 bytes --] [-- Attachment #2: Type: text/plain, Size: 0 bytes --]
Dan Čermák <dan.cermak@posteo.net> writes: > Hi David, > > David Bremner <david@tethera.net> writes: > >> David Bremner <david@tethera.net> writes: >> >>> David Bremner <david@tethera.net> writes: >>> >>> I've sent a couple of patches [1] for things that stood out to >>> me. Please send me anything you think I missed ASAP; I'd like to release >>> the (hopefully) final release candidate tommorow. >> >> I've tagged 0.31~rc2, and uploaded to notmuchmail.org and debian >> experimental. > > I have started seeing consistent test failures with rc2 on ppc64 and > ppc64le: > - on both platforms "get total messages with closed database" from > T568-lib-thread fails I don't see this failure on ppc64el on Debian [0]. Perhaps there is some toolchain difference? > - on ppc64 "comparing existing to exported symbols" from > T360-symbol-hiding fails as well This is more-or-less known [1]. The problem is that ppc64 reports defined functions as 'D' in nm output. At least on ELF architectures, readelf -Ws is more reliable. How many people would using readelf break things for? Currently the test does not run on macOS, afaik. [0]: https://buildd.debian.org/status/architecture.php?a=ppc64el&suite=experimental&buildd=buildd_ppc64el-ppc64el-osuosl-01 [1]: https://buildd.debian.org/status/fetch.php?pkg=notmuch&arch=ppc64&ver=0.31%7Erc2-1&stamp=1598797686&raw=0
Hi
> Here's a lightly edited output from git shortlog to jog your collective
> memories. Please send patches for NEWS for any non-trivial user visible
> changes.
I think I should have sent a news patch for the "unthreaded mode" addition which went into the 0.30 release (sorry about that). What is the best way to add it now?
Best wishes
Mark
Mark Walters <markwalters1009@gmail.com> writes:
> Hi
>
>> Here's a lightly edited output from git shortlog to jog your collective
>> memories. Please send patches for NEWS for any non-trivial user visible
>> changes.
>
> I think I should have sent a news patch for the "unthreaded mode"
> addition which went into the 0.30 release (sorry about that). What is
> the best way to add it now?
>
Maybe add it to the 0.31 NEWS with a note that the change actually
happened in 0.30. I suspect people are more likely to find it if it's in
the most recent release NEWS.
d
[-- Attachment #1.1: Type: text/plain, Size: 1163 bytes --] David Bremner <david@tethera.net> writes: > Dan Čermák <dan.cermak@posteo.net> writes: > >> Hi David, >> >> David Bremner <david@tethera.net> writes: >> >> I have started seeing consistent test failures with rc2 on ppc64 and >> ppc64le: >> - on both platforms "get total messages with closed database" from >> T568-lib-thread fails > > I don't see this failure on ppc64el on Debian [0]. Perhaps there is some > toolchain difference? This specific issue has now disappeared on both ppc64 and ppc64el, so I guess it was just some toolchain breakage that got fixed in the meantime. > >> - on ppc64 "comparing existing to exported symbols" from >> T360-symbol-hiding fails as well > > This is more-or-less known [1]. The problem is that ppc64 reports > defined functions as 'D' in nm output. At least on ELF architectures, > readelf -Ws is more reliable. How many people would using readelf break > things for? Currently the test does not run on macOS, afaik. So it is ok if I just skip this test on ppc64? Or can anything be done about this (I must admit that I have relatively little insight into this)? Cheers, Dan [-- Attachment #1.2: signature.asc --] [-- Type: application/pgp-signature, Size: 865 bytes --] [-- Attachment #2: Type: text/plain, Size: 0 bytes --]
[-- Attachment #1.1: Type: text/plain, Size: 1163 bytes --] David Bremner <david@tethera.net> writes: > Dan Čermák <dan.cermak@posteo.net> writes: > >> Hi David, >> >> David Bremner <david@tethera.net> writes: >> >> I have started seeing consistent test failures with rc2 on ppc64 and >> ppc64le: >> - on both platforms "get total messages with closed database" from >> T568-lib-thread fails > > I don't see this failure on ppc64el on Debian [0]. Perhaps there is some > toolchain difference? This specific issue has now disappeared on both ppc64 and ppc64el, so I guess it was just some toolchain breakage that got fixed in the meantime. > >> - on ppc64 "comparing existing to exported symbols" from >> T360-symbol-hiding fails as well > > This is more-or-less known [1]. The problem is that ppc64 reports > defined functions as 'D' in nm output. At least on ELF architectures, > readelf -Ws is more reliable. How many people would using readelf break > things for? Currently the test does not run on macOS, afaik. So it is ok if I just skip this test on ppc64? Or can anything be done about this (I must admit that I have relatively little insight into this)? Cheers, Dan [-- Attachment #1.2: signature.asc --] [-- Type: application/pgp-signature, Size: 865 bytes --] [-- Attachment #2: Type: text/plain, Size: 0 bytes --]
Dan Čermák <dan.cermak@posteo.net> writes:
> David Bremner <david@tethera.net> writes:
>
>> Dan Čermák <dan.cermak@posteo.net> writes:
>>
>>> - on ppc64 "comparing existing to exported symbols" from
>>> T360-symbol-hiding fails as well
>>
>> This is more-or-less known [1]. The problem is that ppc64 reports
>> defined functions as 'D' in nm output. At least on ELF architectures,
>> readelf -Ws is more reliable. How many people would using readelf break
>> things for? Currently the test does not run on macOS, afaik.
>
> So it is ok if I just skip this test on ppc64? Or can anything be done
> about this (I must admit that I have relatively little insight into
> this)?
I'd say it's OK to skip it for now. Hopefully we will fix it for the
next release; I didn't want to make the switch to readelf in the late
stages of a release.
d
[-- Attachment #1.1: Type: text/plain, Size: 1034 bytes --] Dan Čermák <dan.cermak@posteo.net> writes: > David Bremner <david@tethera.net> writes: > >> Dan Čermák <dan.cermak@posteo.net> writes: >> >>> Hi David, >>> >>> David Bremner <david@tethera.net> writes: >>> >>> I have started seeing consistent test failures with rc2 on ppc64 and >>> ppc64le: >>> - on both platforms "get total messages with closed database" from >>> T568-lib-thread fails >> >> I don't see this failure on ppc64el on Debian [0]. Perhaps there is some >> toolchain difference? > > This specific issue has now disappeared on both ppc64 and ppc64el, so I > guess it was just some toolchain breakage that got fixed in the > meantime. I was wrong, "get total messages with closed database" from T568-lib-thread still fails sometimes on ppc64, but unfortunately not consistently. It appears to mostly fail on a worker with relatively little memory (3584 MB vs 8192 MB or 11264MB). I'll try to keep an eye out on that hoping to find some indicator what could be wrong. Cheers, Dan [-- Attachment #1.2: signature.asc --] [-- Type: application/pgp-signature, Size: 865 bytes --] [-- Attachment #2: Type: text/plain, Size: 0 bytes --]
Dan Čermák <dan.cermak@posteo.net> writes:
>
> I was wrong, "get total messages with closed database" from
> T568-lib-thread still fails sometimes on ppc64, but unfortunately not
> consistently. It appears to mostly fail on a worker with relatively
> little memory (3584 MB vs 8192 MB or 11264MB). I'll try to keep an eye
> out on that hoping to find some indicator what could be wrong.
>
If you have an interactive ppc64 machine (I'm guessing not, but someone
might), it would be interesting to run something like
cd test && count=0 && while ./T568-lib-thread.sh; do
count=$((count + 1));
echo $count
done
And see if the test is just flaky, and fails after some number of
repetitions.
FWIW, it passes 150 times on amd64.
I guess another thing to try would be serialize the test suite on the
low memory machines (set NOTMUCH_TEST_SERIALIZE to some non-empty value,
or remove 'parallel' from the path
d