Re: [PATCH] lib: work around xapian bug with get_mset(0,0, x)
David Bremner writes: > At least Fedora28 triggers this Xapian bug due to some toolchain change . > >https://bugzilla.redhat.com/show_bug.cgi?id=1546162 > > The underlying bug is fixed in xapian commit f92e2a936c1592, and > should be fixed in Xapian 1.4.6 pushed as part of 0.26.2 release ___ notmuch mailing list notmuch@notmuchmail.org https://notmuchmail.org/mailman/listinfo/notmuch
Re: [PATCH] lib: work around xapian bug with get_mset(0,0, x)
David Bremner writes: > At least Fedora28 triggers this Xapian bug due to some toolchain change . > >https://bugzilla.redhat.com/show_bug.cgi?id=1546162 > > The underlying bug is fixed in xapian commit f92e2a936c1592, and > should be fixed in Xapian 1.4.6 pushed to origin and master d ___ notmuch mailing list notmuch@notmuchmail.org https://notmuchmail.org/mailman/listinfo/notmuch
Re: [PATCH] lib: work around xapian bug with get_mset(0,0, x)
David Bremner writes: > David Bremner writes: > >> At least Fedora28 triggers this Xapian bug due to some toolchain change . >> >>https://bugzilla.redhat.com/show_bug.cgi?id=1546162 >> >> The underlying bug is fixed in xapian commit f92e2a936c1592, and >> should be fixed in Xapian 1.4.6 > > pushed to origin and master Ugh. To release and master. d ___ notmuch mailing list notmuch@notmuchmail.org https://notmuchmail.org/mailman/listinfo/notmuch
Re: [PATCH] lib: work around xapian bug with get_mset(0,0, x)
Daniel Kahn Gillmor writes: > But if we want a compile-time check, i don't think we'd need anything > fancier than something like (potentially even directly in query.cc): > > #if XAPIAN_AT_LEAST(1,4,6) > #define MSET_GET_MIN_COUNT 0 > #else > #define MSET_GET_MIN_COUNT 1 > #endif > > what else were you thinking we'd need? > > --dkg I forgot about the XAPIAN_AT_LEAST macro. Nonetheless, I'm not keen to ship untested/uncompiled code (no matter how "obviously correct" it is), so I'd prefer to wait until Xapian 1.4.6 actually exists before adding an ifdef like that. FWIW I doubt that the performance impact of passing 1 instead of 0 here is measurable. For me, getting a bug fix release out is a matter of some urgency; fine tuning and testing against xapian 1.4.6 can wait. d ___ notmuch mailing list notmuch@notmuchmail.org https://notmuchmail.org/mailman/listinfo/notmuch
Re: [PATCH] lib: work around xapian bug with get_mset(0,0, x)
On Wed 2018-04-18 20:13:54 -0300, David Bremner wrote: > Daniel Kahn Gillmor writes: >> On Fri 2018-04-06 08:43:07 -0300, David Bremner wrote: >>> At least Fedora28 triggers this Xapian bug due to some toolchain change . >>> >>>https://bugzilla.redhat.com/show_bug.cgi?id=1546162 >>> >>> The underlying bug is fixed in xapian commit f92e2a936c1592, and >>> should be fixed in Xapian 1.4.6 >> >> is there any way that we can apply this change only if we detect that >> we're running with an unfixed version of Xapian? > > It's possible, but I think we'd need a new configure check, and some > ifdefs. We could consider doing that later, but I don't think it's the > right approach for a bugfix release. I was thinking of a runtime check, not a compile-time check -- to ensure that we drop the workaround as soon as the library is upgraded beneath us. But if we want a compile-time check, i don't think we'd need anything fancier than something like (potentially even directly in query.cc): #if XAPIAN_AT_LEAST(1,4,6) #define MSET_GET_MIN_COUNT 0 #else #define MSET_GET_MIN_COUNT 1 #endif what else were you thinking we'd need? --dkg ___ notmuch mailing list notmuch@notmuchmail.org https://notmuchmail.org/mailman/listinfo/notmuch
Re: [PATCH] lib: work around xapian bug with get_mset(0,0, x)
Daniel Kahn Gillmor writes: > On Fri 2018-04-06 08:43:07 -0300, David Bremner wrote: >> At least Fedora28 triggers this Xapian bug due to some toolchain change . >> >>https://bugzilla.redhat.com/show_bug.cgi?id=1546162 >> >> The underlying bug is fixed in xapian commit f92e2a936c1592, and >> should be fixed in Xapian 1.4.6 > > is there any way that we can apply this change only if we detect that > we're running with an unfixed version of Xapian? > > --dkg It's possible, but I think we'd need a new configure check, and some ifdefs. We could consider doing that later, but I don't think it's the right approach for a bugfix release. d ___ notmuch mailing list notmuch@notmuchmail.org https://notmuchmail.org/mailman/listinfo/notmuch
Re: [PATCH] lib: work around xapian bug with get_mset(0,0, x)
On Fri 2018-04-06 08:43:07 -0300, David Bremner wrote: > At least Fedora28 triggers this Xapian bug due to some toolchain change . > >https://bugzilla.redhat.com/show_bug.cgi?id=1546162 > > The underlying bug is fixed in xapian commit f92e2a936c1592, and > should be fixed in Xapian 1.4.6 is there any way that we can apply this change only if we detect that we're running with an unfixed version of Xapian? --dkg ___ notmuch mailing list notmuch@notmuchmail.org https://notmuchmail.org/mailman/listinfo/notmuch
Re: [PATCH] lib: work around xapian bug with get_mset(0,0, x)
On Fri, Apr 06 2018, David Bremner wrote: > At least Fedora28 triggers this Xapian bug due to some toolchain change . > >https://bugzilla.redhat.com/show_bug.cgi?id=1546162 > > The underlying bug is fixed in xapian commit f92e2a936c1592, and > should be fixed in Xapian 1.4.6 > --- > > I have verified this doesn't break the test suite in my environment, > but I would appreciate feedback from someone with a Fedora 28 test > platform. Without this change, on Fedora 28 container: 9 aborts. test execution halts at T570-revision-tracking.2 since ${result} is empty string with this change: failure count is about the same as with fedora 27 so, +1 That, is my setup, pulled fedora:28 docker image and run a script to install some stuff there. I'd bet I can get the error count down by tuning the creation script... Tomi ___ notmuch mailing list notmuch@notmuchmail.org https://notmuchmail.org/mailman/listinfo/notmuch