Re: [Bacula-devel] Bacula Director reload race condition/crash

2012-06-11 Thread Martin Simmons
The code is broken (not thread-safe). Try moving the locks furthur out, i.e. move lock_jobs(); LockRes(); before if (already_here) { and move UnlockRes(); unlock_jobs(); after already_here = false; __Martin > On Sat, 09 Jun 2012 11:55:43 +1000, Alexander Wigen said:

Re: [Bacula-devel] Regression tests - compile failures

2012-06-11 Thread Dan Langille
On 2012-06-09 03:17, Kern Sibbald wrote: > Interesting. > > Unfortunately, we cannot use changes to configure because it is > "compiled" > from autoconf/configure.in by the autoconf tools. Understood. But > > I just checked the ./configure of the latest code that I built last > week > on 8.2 a

[Bacula-devel] Git repo

2012-06-11 Thread Kern Sibbald
Hello, In looking at how to back out a bad community patch that I put into Bacula 5.2.8, I decided to reset the repo and to prior to that patch, because for me it was by far the simplest thing to do. The consequence is that for any of you who checked out or updated the database after 5.2.8 was re

Re: [Bacula-devel] Regression tests - compile failures

2012-06-11 Thread Kern Sibbald
On 06/11/2012 05:01 PM, Dan Langille wrote: > On 2012-06-11 08:11, Dan Langille wrote: >> On 2012-06-09 03:17, Kern Sibbald wrote: >>> Interesting. >>> >>> Unfortunately, we cannot use changes to configure because it is >>> "compiled" >>> from autoconf/configure.in by the autoconf tools. >> >> Und

Re: [Bacula-devel] Regression tests - compile failures

2012-06-11 Thread Dan Langille
On 2012-06-11 08:11, Dan Langille wrote: > On 2012-06-09 03:17, Kern Sibbald wrote: >> Interesting. >> >> Unfortunately, we cannot use changes to configure because it is >> "compiled" >> from autoconf/configure.in by the autoconf tools. > > Understood. > > But > Sorry, I forgot to finish that. W

Re: [Bacula-devel] Regression tests - compile failures

2012-06-11 Thread Martin Simmons
Looks like this problem: http://lists.gnu.org/archive/html/bug-autoconf/2010-08/msg00071.html caused by libtool defining ECHO, which clashes with Bacula's definition. I see that there was a similar problem with RM (renamed to REMOVE), but the simplest fix is probably to delete the ECHO variable

Re: [Bacula-devel] Regression tests - compile failures

2012-06-11 Thread Kern Sibbald
On 06/11/2012 01:49 PM, Martin Simmons wrote: > Looks like this problem: > > http://lists.gnu.org/archive/html/bug-autoconf/2010-08/msg00071.html > > caused by libtool defining ECHO, which clashes with Bacula's definition. > > I see that there was a similar problem with RM (renamed to REMOVE), but

[Bacula-devel] Bacula release 5.2.9

2012-06-11 Thread Kern Sibbald
Hello, As some of you will know, we recently released version 5.2.7 followed a week later by 5.2.8, and now a day later by 5.2.9. Both 5.2.7 and 5.2.8 had some minor glitches, mostly involving Sparc and platforms other than Linux. I believe they are all resolved now. So, we have released Bacula

Re: [Bacula-devel] Regression tests - compile failures

2012-06-11 Thread Martin Simmons
> On Mon, 11 Jun 2012 11:01:30 -0400, Dan Langille said: > > What FreeBSD system are you working with? I'm asking in case there is > something in your system > which is doing something different. FWIW, GNU make seems to silently ignore the malformed line and I always use that to compile Bac

Re: [Bacula-devel] Regression tests - compile failures

2012-06-11 Thread Kern Sibbald
On 06/11/2012 09:27 PM, Martin Simmons wrote: >> On Mon, 11 Jun 2012 11:01:30 -0400, Dan Langille said: >> What FreeBSD system are you working with? I'm asking in case there is >> something in your system >> which is doing something different. > FWIW, GNU make seems to silently ignore the malf

Re: [Bacula-devel] Bacula Director reload race condition/crash

2012-06-11 Thread Alexander Wigen
Hi Martin, I have confirmed your suggestion works. The patch tested is attached. Will this fix be accepted into the public release, if not what is the way forward? Cheers, Alex On 11/06/12 21:19, Martin Simmons wrote: The code is broken (not thread-safe). Try moving the locks furthur out,