Hello, The direct answer to your question from email subject is because a code was touching a memory which was freed, as you can spot in this line of the log:
#3 VOLRES::set_slot (slot=-1, this=0x5555555555555555) at vol_mgr.h:88 The pointer address for "this" was read from memory which was already released (freed) and reads intentionally made "garbage" - it is the 0x5555... value made for (AFAIR, correct me if I've mistaken. If I'm wrong then it is an "uninitialized" memory access as this 0x5555... is intentionally put by the bacula core libs to spot such issues). pon., 23 sty 2023 o 15:12 Chandler Sobel-Sorenson <ad...@genome.arizona.edu> napisał(a): > I'll admit I'm using obsolete versions of the bacula demons, but they have > been working great. I know it would be good to upgrade, I try to keep > everything in my realm upgraded, but with my limited time and varying > versions of OSes, it's better to keep using the old version, still. > > All I did was make some small changes to: baconfig.h, dird/dird_conf.c, > filed/restore.c, and findlib/create_file.c. You should share what you have changed to be able to confirm if your changes caused the issue. > I suppose the changes to baconfig.h triggered pretty much everything to be > re-compiled including bacula-sd. That leads me to think it's crashing due > to the newer compiler or OS being used? These binaries were last compiled > on Debian 8 using GCC 4. Now I'm on Debian 10 with GCC 8. The mods I made > were successful, and bacula-dir and bacula-fd start and run fine. For some > reason, bacula-sd crashes with the below debug output plus traceback, but > none of the modified files are mentioned. Instead, you can see the debug > code leading up the crash. > It is not how the compiler and source code is working. You've got a crash at VOLRES::set_slot() method which does not mean it is a source of the issue. The issue is an access to already freed memory which brings a main question - why a memory was freed before access? Where did this memory free happen? Is it ok that a memory was freed there? etc. etc. > Fortunately, I saved backups of the binaries before running `make install` > and the backup bacula-sd still runs fine with the newly compiled bacula-dir > and -fd, although I'm not sure it's a good idea since it's based on the > previous version of baconfig.h. I'd like to know why the newly compiled > version is suddenly crashing, can you make any sense of the below output? > We know why it is crashing (as explained above), but the root cause of why it ended up in this situation is unknown. You have a dozen possibilities including your changes. best regards -- Radosław Korzeniewski rados...@korzeniewski.net
_______________________________________________ Bacula-devel mailing list Bacula-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-devel
