[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#511011: libtdb1: Breaks ABI without SONAME bump.



Cyril Brulebois <kibi@debian.org> (06/01/2009):
> Package: libtdb1
> Version: 1.1.2~git20080615-1
> Severity: serious
> Justification: ABI breakage.

Stripping some parts for -release, keeping what matters:

> We don't care about 'U' ones, of course. Checking which ones got away:
> | grep ^-T tdb-symbols.diff
> | -T tdb_clear_spinlocks
> | -T tdb_create_rwlocks
> | -T tdb_logging_function
> | -T tdb_set_lock_alarm
> | -T tdb_spinlock
> | -T tdb_spinunlock
> 
> That *might* not be the problem if they are not exposed through the API.
> But both tdb_logging_function and tdb_set_lock_alarm are exposed.
> 
> I'm going to have a look at the packages that depend on this library, to
> see how things looks like. I'm in contact with a release manager as well
> (hello dato).

Here is my findings about etch/lenny packages WRT the 6 removed symbols
(looking at main).

Packages both in etch and in lenny:
-----------------------------------
libprintsys/etch:   OK (basic symbols)
libprintsys/lenny:  OK (basic symbols)
pconf-detect/etch:  OK (linked in, no symbols)
pconf-detect/lenny: OK (linked in, no symbols)

Packages only in lenny:
-----------------------
fdm/lenny:        OK (only a few symbols)
python-tdb/lenny: OK (lots of symbols, built from the same source)

Packages only in etch:
----------------------
adept-installer/etch:    OK (linked in, no symbols)
adept-manager/etch:      OK (linked in, no symbols)
adept-notifier/etch:     OK (linked in, no symbols)
adept-updater/etch:      OK (linked in, no symbols)
libprinterconf0c2a/etch: OK (linked in, no symbols)
tagcolledit/etch:        OK (linked in, no symbols)
tdb-tools/etch:          OK (lots of symbols)


Note that it still can break locally-built packages or pieces of
software. I think you could consider downgrading the severity, or
tagging it lenny-ignore since there seem to be no collateral damages
within Debian.

That said, I guess a t-p-u upload to fix #510886 and a binNMU of fdm
will do the job (I didn't check yet it needs a >= on the lenny version,
but I guess it'd be a a nice-to-have anyway).

Mraw,
KiBi.

Attachment: signature.asc
Description: Digital signature


Reply to: