Raphael Hertzog writes ("Re: Bug#143307: triggers-related dependtry assertion
fix"):
> Can you expand on why you decided to create a msdbrw_simulate status then
> as opposed to using modstatdb_note_ifwrite() in the problematic cases
> in the trigger code ?
Various other bits of the code need to know the simulation status too
- for example, the triggers code mustn't think it has incorporated
triggers permanently and remove them from Deferred. This was most
easily achieved by encoding it in the msdbrw.
Also, to be honest, I think I didn't like the way I had done it
before. Having an explicitly simulating mode for the dbmodify
machinery seems to make more sense, particularly as that machinery has
gained more semantic knowledge and its data is spread across more
files.
> And why didn't you update the check in modstatdb_note_ifwrite() as well ?
This was a mistake.
For the two calls in question, in triglib.c, there is no actual
difference in the current code except that _ifwrite won't do the
logging and status fd reporting, which we _do_ want done.
So _ifwrite should say
if (cstatus >= msdbrw_simulate)
(I don't think this is a particularly important case but it's also
pretty harmless to push into a stable tree now.)
Ian.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
