Hi Huidae,
(cc to "grass5")
On Tue, Aug 07, 2001 at 12:07:08PM +0900, GRASS wrote:
Hi Markus,
> Hi Huidae,
> I have seen you have been working on error.c. Since a few weeks
> the presence of $HOME/GIS_ERROR_LOG doesn't avoid sending GRASS
> error mails any more. In former versions the user could avoid the
> mails after invoking the command:
>
> touch $HOME/GIS_ERROR_LOGWell, I understood that GRASS_STDERR env var should be set to avoid mailing.
So only the presence of $HOME/GIS_ERROR_LOG doesn't help.
oh - a new idea 
Well, how to set the GRASS_STDERR? Something like
export GRASS_STDERR=/home/neteler/GIS_ERR
doesn't seem to work.
>
> In my opinion the idea of the function seems a bit unusual:
>
> src/libes/gis/error.c
> * G_warning (msg)
> * char *msg
> *
> * Gives the message, msg, to the user. Also print a message to
> * $GISBASE/GIS_ERROR_LOG if the file exists and is writeable;
> ^^^^^^^^^-> this should be HOME I think
> * and to the file $HOME in the users home directory if $HOME is set.
> * G_warning() returns, which G_fatal_error() exits.
>
> usually users don't have write access in $GISBASE, so I suggest
> to change it to $HOME/GIS_ERROR_LOG
>You're right about that. However, we can not know the fully qualified
mapset from the current file format. So what about putting GIS_ERROR_LOG
into its own mapset directory(not PERMANENT), which is also writable by
users. Furthermore, I think that the format should include mapset info.If users don't have write access in $GISBASE at all, $HOME is ok. But in
this case, mapset info is inevitable in a log file.> What do you think?
In fact, I've never known this feature before you asked me.
Yours,
Huidae Cho
I suggest to discuss above here in grass5 list. Any opinions?
Markus