Builder libpcap-solaris10-sparc Build #3242

Results:

Build successful

SourceStamp:

BuildSlave:

Reason:

The SingleBranchScheduler scheduler named 'schedule-libpcap-solaris10-sparc' triggered this build

Steps and Logfiles:

1. git update ( 27 secs )
   1. stdio
2. shell Set developer mode ( 0 secs )
   1. stdio
3. shell_1 './autogen.sh' ( 25 secs )
   1. stdio
4. shell_2 './configure' ( 1 mins, 10 secs )
   1. stdio
   2. config.log
5. shell_3 'gmake' ( 56 secs )
   1. stdio
6. shell_4 'gmake check' ( 16 mins, 55 secs )
   1. stdio
7. shell_5 'rm -rf ...' ( 0 secs )
   1. stdio
8. shell_6 'gmake install ...' ( 5 secs )
   1. stdio
9. trigger triggered trigger-tcpdump-solaris10-sparc ( 16 mins, 43 secs )
   1. - no logs -
   2. tcpdump-solaris10-sparc #6166

Build Properties:

Forced Build Properties:

Responsible Users:

1. Guy Harris

   gharrisohnoyoudont@sonic.net

Timing:

All Changes:

:

1. Change #256915

   Category	libpcap
   Changed by	Guy Harris <gharrisohnoyoudont@sonic.net>
   Changed at	Tue 03 Feb 2026 21:26:11
   Repository	https://git.tcpdump.org/libpcap
   Project	libpcap
   Branch	master
   Revision	b8d160a250ed87916f4dfe5671ee90bbb75dc2cf

   Comments

   Free p->opt.device on close, not on cleanup.
   That way, if a failed pcap_activate() cleans up anything done by the
   activate by calling the module's cleanup routine, which should always
   call pcapint_cleanup_live_common(), p->opt.device, which is set by
   pcap_create() and shouldn't be freed or modified until the pcap_t is
   closed, will be left alone.
   
   Commit e333a6044f7d2d3225a6a22205b6b7c1e389945f moved the freeing
   of p->opt.device to pcap_cleanup_live_common(), but the reason for doing
   that is that, before that commit, we were freeing p->opt.device *before*
   calling the module's cleanup routine, and pcap-npf.c's cleanup routine
   used p->opt.device to turn monitor mode off.
   
   This change moves freeing of p->opt.device back, but does it *after* the
   module's cleanup routine is called, so it's safe for that rouine to use
   p->opt.device - as long as it doesn't save the value of p->opt.device
   and attempt to use it later. Copy the string if you need to do that.
   
   Fix #1615.
   
   Add a test for that crash in testprogs/activatetest.c. That won't
   necessarily catch *all* cases where this is a problem on *all*
   platforms, but it should at least catch it on Linux.

   Changed files

   • pcap.c
   • testprogs/activatetest.c