GNU bug report logs - #33210
Cuirass: Use a SQLite in single-thread mode

Previous Next

Full log

View this message in rfc822 format

From: Clément Lassieur <clement <at> lassieur.org>
To: Danny Milosavljevic <dannym <at> scratchpost.org>
Cc: 33210 <at> debbugs.gnu.org
Subject: [bug#33210] Cuirass: Use a SQLite in single-thread mode
Date: Tue, 06 Nov 2018 15:07:40 +0100

Hey Danny,

Danny Milosavljevic <dannym <at> scratchpost.org> writes:

> Hi Clément,
>
> On Tue, 06 Nov 2018 01:50:11 +0100
> Clément Lassieur <clement <at> lassieur.org> wrote:
>
>> > This is not supposed to happen in relational database systems.  The reason
>> > why it happens here is because we use the same transaction (session) for
>> > both the updates done by the evaluator and the queries done by the web
>> > interface.  It would be much better if the queries by the web interface
>> > had their own transaction.  It was fine before but in some refactoring,
>> > "evaluate" ceased to be an external program and I didn't notice what
>> > happened to it.  
>> 
>> Yes, I know, but I remember that I told you[1] that there was an easy
>> (one line) fix for this. :-)
>
> That would really only be a workaround (arguably still better than what we
> have now - which is basically you ask for a banana and I give you an apple
> - what if other procedures in cuirass assume it's a banana? :) ).
>
> We would be doing the work that SQLite would have done, but we'd do it in a bad
> way (by blocking everyone else).
>
> It's not really useful to undo all the progress relational databases have made.
>
> If we had multiple transactions in progress touching the same row, SQLite would
> use MVCC to hold diverging copies of the same row at the same time, blocking
> nobody.

It blocks everyone indeed, because we take care of the scheduling in a
rather basic way.  But if I understand correctly, the overall spent time
is more or less the same: only the order of the requests differs.  There
is an essential difference however: if we take care of the scheduling,
we won't have SQLITE_BUSY blocking the Fibers scheduler all the time.
And blocking the Fibers scheduler has an impact on all other possibly
unrelated Fibers clients.  When guile-sqlite3 handles SQLITE_BUSY
correctly, I'll be happy to switch back to a multi-threading SQLite.
While waiting for it, I believe our users want a fast Cuirass, and I'd
like the time spent in the Fibers scheduler to be balanced by removing
the SQLite now useless mutexes.

Does that make sense?

Clément

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.