[Rpm-maint] Re: [Suse patch] Don't block signals in ro-mode

[Panu Matilainen]

On Fri, 25 May 2007, Michael Schroeder wrote:

> On Fri, May 25, 2007 at 01:47:46PM +0300, Panu Matilainen wrote:
>> On Fri, 25 May 2007, Panu Matilainen wrote:
>>>
>>> With this patch, stale ro-locks are left behind when running as root and
>>> interrupted (eg rpm -qa), not good.
>
> Oh, so that's why the signals are also blocked in ro mode. I always
> wondered if that was just a mistake.
>
> So the signal handling patch should check if the db is private
> or not.
>
> The big question is, of course, if you want to support both
> locking strategies or not.

That's of course a big question - one that I'd rather not think about 
right now.

But actually for queries, I think the big question is why isn't the db 
opened in fully readonly mode in the first place, but in joinenv mode if 
the directory is writable? To allow lazy initialization of the db if it 
doesn't exist I suppose, but...

>> ...unless private locking is used, which is what Suse does (patch below).
>> Isn't this already configurable through %__dbi_foo macros?
>
> Yes, this patch is probably not really needed. I was just too
> lazy to check the code. ;-)

Heh :)

 	- Panu -