Gavin Sherry <swm(at)linuxworld(dot)com(dot)au> writes:
> This small patch reloads dynamic libraries whose modification time is
> greater than that at the time it was initially loaded. This means that
> connections do not need to be reinitialised when a library is recompiled.
Is that a good idea? It's easy to imagine cases where a library is not
designed to be unloaded (eg, it hooks into things in the main backend
and doesn't have a way to unhook). I'd rather stick with the current
behavior of unload/reload only when specifically commanded to.
The patch as given fails in the same inode/different path case, btw.
I think you wanted to make the test further down.
regards, tom lane
In response to
pgsql-hackers by date
|Next:||From: Tom Lane||Date: 2002-05-04 16:18:24|
|Subject: Re: Native Windows, Apache Portable Runtime |
|Previous:||From: Tom Lane||Date: 2002-05-04 15:20:39|
|Subject: Re: Using views and MS access via odbc |
pgsql-patches by date
|Next:||From: Manfred Koizar||Date: 2002-05-05 15:29:24|
|Subject: Reduce per tuple overhead (bitmap)|
|Previous:||From: Gavin Sherry||Date: 2002-05-04 11:05:35|
|Subject: Auto-reload of dynamic libraries|