Re: [gamin] various regressions in gamin HEAD CVS

[John McCutchan <ttb tentacle dhs org>]

On Mon, 2005-08-22 at 17:51 +0200, Frederic Crozat wrote:
> Hi,
> 
> I've discovered various regressions in gamin CVS HEAD, compared to 0.1.5
> release :
> 
> -debug output is no longer enabled by default, unless you are building
> CVS. I think it is wrong, because it was quite useful even for released
> code, for packagers, at least until all possible regressions are catched
> (I know it can be enabled back).
> 

Do you mean it isn't compiled with -g unless you are compiling from CVS?
If so, I agree. I will make that change. Otherwise, please be more
specific. 

> -even when enabled explicitly, it seems GAM_DEBUG variable doesn't have
> any effect and no output is sent at all.
> 

I made a change to that code the other day. I thought I was making sure
that if GAM_DEBUG is set that we will get debug output. I tested it, to
make sure that it did work, I will double check this today.

> -when forcing dnotify backend using GAM_TEST_DNOTIFY, I got a crash when
> removing and re-creating a bunch of files rapidly :
> GAM_TEST_DNOTIFY=1 /home/gnome/prefix212/libexec/gam_server --notimeout
> test
> 
> ** ERROR **: file gam_poll_generic.c: line 119
> (gam_poll_generic_remove_missing): assertion failed: (g_list_find
> (missing_resources, node))
> aborting...

Sorry, I added this bug. I was trying to be stricter about when
remove_missing/busy is called. It turned out to be cleaner if we just
allowed them to be called even if the resource wasn't on the
missing/busy list. I thought I got rid of the assertions, this is one I
missed. It will be gone shortly.

-- 
John McCutchan <ttb tentacle dhs org>