Bug#202746: gcc: -O2 causes bogutil to segfault

[Matthias Andree]

Clint Adams <schizo at debian.org> writes:

[stripping the other recipients]

>> > - can you identify the file beeing miscompiled?
>> 
>> No, I don't seem to be able to do that.  If I compile with -O1, it
>> works.  If I compile with -O2, it segfaults.  If I compile all objects
>> -O2 except the ones that seem to be relevant in the backtrace, it still
>> segfaults.  (Also tried with various random .c files that didn't seem
>> relevant).
>
> I take that back.  If all files are compiled with -O2 except for
> datastore_db.c (compiled with -O0), it works.

We've had a stable datastore_db.c until July 15 (rev. 1.21), since July
16 there have been some changes (7 versions, up to rev. 1.28). Can you
identify the version which introduced the failure? We may then be able
to cut the patch from the "good" to the "broken" version in halves
(divide-and-conquer style) and figure the failure-inducing
change. (Maybe you need to revert to an older bogofilter version and
gradually try the newer datastore_db.c versions.)

I haven't been able to audit datastore_db.c changes and will not be able
to for a couple of days. Being at it, I find that datastore_db.c has
grown too much, and we may need to separate the database library (how do
we obtain key/value pairs) interface from the data representation (how
is the value field structured (and what do we use for
key)). Unfortunately, I won't be able to do that for some days,
i. e. not before some days into August. :-(

-- 
Matthias Andree