Builder libpcap-solaris11-amd64 Build #200

Results:

Exception

SourceStamp:

BuildSlave:

Reason:

The SingleBranchScheduler scheduler named 'schedule-libpcap-solaris11-amd64' triggered this build

Steps and Logfiles:

1. git update ( 7 secs )
   1. stdio
2. shell './build_matrix.sh' ( 7 hrs, 53 mins, 33 secs )
   1. stdio
   2. config.log
3. trigger triggered trigger-tcpdump-solaris10-amd64 ( 3 secs )
   1. - no logs -

Build Properties:

Forced Build Properties:

Responsible Users:

1. Denis Ovsienko

   denisohnoyoudont@ovsienko.info

Timing:

All Changes:

:

1. Change #285330

   Category	libpcap
   Changed by	Denis Ovsienko <denisohnoyoudont@ovsienko.info>
   Changed at	Fri 09 Oct 2026 16:26:33
   Repository	https://git.tcpdump.org/libpcap
   Project	libpcap
   Branch	master
   Revision	a8fce18e938c7d91874f0ec718293dbfa5b67914

   Comments

   Make more register handling bugs early failures.
   Introduce new_arth() to fail if an arithmetic expression is going to be
   used without allocating a register first.  Introduce alloc_reg_arth() to
   fail when for the same arithmetic expression a new allocation is about
   to leak an existing allocation.  Use the new functions in gen_arth(),
   gen_loadlen() and gen_loadi_internal().

   Changed files

   • gencode.c
   • testprogs/TESTrun

2. Change #285331

   Category	libpcap
   Changed by	Denis Ovsienko <denisohnoyoudont@ovsienko.info>
   Changed at	Fri 09 Oct 2026 18:42:23
   Repository	https://git.tcpdump.org/libpcap
   Project	libpcap
   Branch	master
   Revision	f1da34eb2c99b0f61932dc6d9274b4acacdb76b5

   Comments

   Reduce register footprint of gen_load_internal().
   Allocate the output scratch memory register just before using it and
   after deallocating the input register, rather than at the beginning.
   
   This way it is still possible to run out of registers before
   gen_load_internal() or downstream of it in gen_abs_offset_varpart(), but
   not when allocating the output register anymore, also one more register
   is available in gen_abs_offset_varpart() than before, which in corner
   cases that involve "proto[expr]" allows to compile slightly more complex
   expressions or to start accessing packet data using an offset that has a
   variable part, although not both at the same time.
   
   Consequently, four filter reject tests become filter accept tests and
   only the 16 register plane (in the sense of commit 17c2d1e) comprises
   both accept and reject tests.  Update many other filter accept tests
   that now use a lower index for one of the registers as a result of this
   change.

   Changed files

   • gencode.c
   • testprogs/TESTrun

3. Change #285332

   Category	libpcap
   Changed by	Denis Ovsienko <denisohnoyoudont@ovsienko.info>
   Changed at	Fri 09 Oct 2026 18:42:23
   Repository	https://git.tcpdump.org/libpcap
   Project	libpcap
   Branch	master
   Revision	d5d932c87067018f74c945c2b70ecc5dec725dc5

   Comments

   TESTrun: Add more scratch memory capacity tests.
   Require both gen_loadi_internal() and gen_loadlen() to allocate exactly
   one register and not to leak it.

   Changed files

   • testprogs/TESTrun

4. Change #285333

   Category	libpcap
   Changed by	Denis Ovsienko <denisohnoyoudont@ovsienko.info>
   Changed at	Sat 10 Oct 2026 23:05:26
   Repository	https://git.tcpdump.org/libpcap
   Project	libpcap
   Branch	master
   Revision	af7ef0678be69b157bc988743a1eae72a212b58a

   Comments

   BPF_MEMWORDS actually counts registers. [skip ci]

   Changed files

   • pcap/bpf.h