On Tue, Sep 24, 2013 at 9:36 PM, Richard Hipp <d...@sqlite.org> wrote:
> I think this is a bug. My theory is that Fossil is putting itself in a
> chroot jail to handle each request, but it is doing so after opening the
> database file. That's a huge no-no. The database file needs to be opened
> after entering the chroot jail. If the database file is open prior to
> entering the chroot jail, it needs to be closed prior to entry then
> reopened afterwards.
>
enter_chroot_jail() closes/re-opens the repo, but it closes after doing the
chdir. Do i understand correctly that the db_close() should be done before
the chdir, and the db_open_repo...() afterwards?
@Krister: please try the following as a quick-hack fix, and if this works
then i'll refactor this into something nicer:
in main.c:
enter_chroot_jail() (currently line 1156)
at the bottom of the function is this block:
if( g.db!=0 ){
db_close(1);
db_open_repository(zRepo);
}
It looks like:
http://fossil-scm.org/index.html/artifact/0fa850420a5917a8f83e127caf849a2f6cb2f77a?ln=1171-1174
_Copy_ that block to just after the variable declarations (line 1143),
remove the db_open_xxx() part from the first copy and the db_close() part
from the second copy. Then please give it a try.
:-?
--
----- stephan beal
http://wanderinghorse.net/home/stephan/
http://gplus.to/sgbeal
_______________________________________________
fossil-users mailing list
fossil-users@lists.fossil-scm.org
http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users
