Hi Damyan, Damyan Ivanov <d...@debian.org> (2013-05-28): > This all happened, and firebird2.5 version 2.5.2.26540.ds4-1 had 20 > days in unstable without new bugs. It migrated to testing after a > giveback on sparc (first attempt crashed the compiler). Perhaps it is > time to consider it for 7.0.1?
looking at it for 7.2 now, sorry it took so long. > The last patch, out/crash-create-db-restricted.patch is for a crash > that is claimed fixed in upstream Svn, but the actual commit is > missing. I have produced it by looking at the surrounding code. > Upstream bug is at http://tracker.firebirdsql.org/browse/CORE-3996 Given the current status on this page, I suspect a proper commit is available now, I guess you could double check your fix is alright. Some more comments inline. > diff --git a/debian/NEWS b/debian/NEWS > new file mode 100644 > index 0000000..2d884bb > --- /dev/null > +++ b/debian/NEWS > @@ -0,0 +1,16 @@ > +firebird2.5 (2.5.2.26540.ds4-1) unstable; urgency=low > + > + Important for big-endian server installations: > + > + It has been discovered that before Firebird 2.5.2 (as packaged in debian > + package version 2.5.2.26539.ds4-1), CHAR_TO_UUID and UUID_TO_CHAR > + built-in functions work incorrectly on big-endian servers. On such > machines, > + bytes/characters are swapped and go in wrong positions when converting. The > + bug is fixed in this release, but that means these functions now return > + different values than before for the same input parameter. > + > + Additionally, the CHAR_TO_UUID2/UUID_TO_CHAR2 functions that were added as > + fixed variants of the buggy functions in the 2.5.2 development cycle are > + dropped in this release. > + > + -- Damyan Ivanov <d...@debian.org> Wed, 07 Nov 2012 17:52:10 +0200 Versions aren't consistent, maybe that was due to an intermediate package, or maybe you're just trying to get a lower version than what's in jessie/sid? If you're rebuilding the 2.5.2.26540.ds4-1 package for wheezy, you probably should be adding a changelog entry on top of it, using 2.5.2.26540.ds4-1~deb7u1 as a version number to indicate this is a mere rebuild in wheezy (when adding stuff to a given package in wheezy, we usually use $version_in_stable+deb7u1 instead). From a quick glance at the -1 → -4 changelog, it doesn't look like there should be more things you'd like to consider for wheezy, so -1 should still be the package to be "stable-backported"? > diff --git a/debian/changelog b/debian/changelog > index d46c066..eb0b227 100644 > --- a/debian/changelog > +++ b/debian/changelog > @@ -1,3 +1,63 @@ > +firebird2.5 (2.5.2.26540.ds4-1) unstable; urgency=low > + > + * Official 2.5.2 release > + + CORE-3912: segfault in superclassic (Closes: #693192) > + + Restored the on-disk-structure compatibility with 2.5.1 index keys > + (Closes: #693193) > + + Fixed broken (working as no-op) sweep in SuperServer (Closes: #693195) > + + CORE-3902: Derived fields may not be optimized via an index > + (Closes: #693196) > + + CORE-3895: High memory usage when PSQL code SELECT's from stored > + procedure which modified some data (Closes: #693202) > + + CORE-3238: GEN_UUID returns a non-RFC-4122-compliant UUID > + (Closes: #693207) > + + CORE-3887: CHAR_TO_UUID and UUID_TO_CHAR works different in big endian > + architectures (Closes: #693209) > + + Enabled per-table runtime stats for sweeper > + + Changes not concerning Debian > + - CORE-3786: Hangs on MacOSX 10.7 (Lion) on DB create after reboot > + - CORE-3911: API entrypoints Bopen and BLOB_open are not visible on > Darwin > + - CORE-3740: SELECT using IN list with >413 elements causes crash on > Mac > + (stack overflow with default stack size) > + - CORE-3740: optimisation bug in GCC on Darwin Some bug reports above aren't marked in the BTS as affecting stable; it would be nice to add the relevant "found" versions there if you get a chance. > + * Update debian/copyright (two new files, no licensing changes) ACK. > + * Add NEWS.Debian about incompatible fix in char↔UUID conversion functions Joys of behavorial changes due to correctness fixes. :/ I guess that fixing and announcing it through NEWS.Debian is the best we can do… > + * drop patches included in the upstream release: > + + upstream/r54702-cve-2012-5529.patch > + + upstream/r57728-cve-2013-2429.patch > + * refresh separate-file-and-sem-perms.patch to apply cleanly ACK. > + * Patches taken from upstream SVN > + + r57516r57773-gbak-y-redirection.patch > + make gbak -y work with redirection again (regression from 2.5.1) > + http://tracker.firebirdsql.org/browse/CORE-3995 > + + r57707r57710-lots-autonomous-trx-leaks-crash.patch > + fix engine crash/memory leak with many autonomous transactions (remote > + crash/memory leak) > + http://tracker.firebirdsql.org/browse/CORE-3908 > + + r57349-bad-trn-num-logged-during-sweep.patch > + fix invalid transaction counters logged during sweep (trivial fix) > + http://tracker.firebirdsql.org/browse/CORE-3978 > + + r57714r57716-fix-isql-edit-command.patch > + fix isql's edit command broken in 2.5.2 (regression from 2.5.1) > + http://tracker.firebirdsql.org/browse/CORE-3990 > + + r58004-crash-converting-overscaled-numeric-to-string.patch > + fix engine crash while converting an overscaled numeric to a string > + (remote crash) > + http://tracker.firebirdsql.org/browse/CORE-4093 > + + r57795-crash-storing-long-incompressible-data.patch > + fix bugcheck/corruption when storing long uncompressible data (possible > + db corruption) > + http://tracker.firebirdsql.org/browse/CORE-4036 I'll take advantage from the extended testing those patches got during the past months, and from the fact those were picked from/accepted by upstream to convince myself they are OK for stable. > + * add out/crash-create-db-restricted.patch > + fixes a server crash when attempting creation of a database outside of > + allowed paths with firebrd.xonf setting of 'DatabaseAccess' other than > + 'Full' firebird.conf I think; the patch headers seem to confirm that. > + -- Damyan Ivanov <d...@debian.org> Thu, 09 May 2013 16:39:17 +0300 Not tagging this bug report as confirmed for now, as I'd like to get your feedback about versioning first. Besides that particular point, the proposed plan looks good to me. Mraw, KiBi.
signature.asc
Description: Digital signature
