[PATCH 0/8] Remove obstack allocation of hash tables

[Tom Tromey <tromey@adacore.com>]

Obstack allocation of hash tables has slightly irritated me for a long
time.  The main issue with these is that if the hash table is resized,
then the old table is not freed, but instead left around as unused
garbage.

This series removes this capability entirely.  IMO there's no reason
to ever do this -- with C++, heap allocation is just as convenient,
and doesn't cause effective memory leaks.

While working on this I found a few other minor cleanups to do.  This
series includes them all.

Regression tested on x86-64 Fedora 38.

---
Tom Tromey (8):
      Don't obstack-allocate the DIE hash
      Don't obstack-allocate the CU dependency hash table
      Rename symtab::fullname
      Make symtab members private
      Add compunit_symtab::forget_cached_source_info
      Don't obstack-allocate the call site hash table
      Remove hashtab_obstack_allocate
      Prefer htab_traverse_noresize

 gdb/annotate.c             |  2 +-
 gdb/completer.c            |  2 +-
 gdb/dwarf2/cu.c            | 13 +++++-----
 gdb/dwarf2/cu.h            |  6 ++---
 gdb/dwarf2/read.c          | 63 ++++++++++++++++------------------------------
 gdb/gdb_bfd.c              |  2 +-
 gdb/objfiles.c             |  2 ++
 gdb/python/py-breakpoint.c |  4 +--
 gdb/source.c               | 21 ++++++++--------
 gdb/symfile-debug.c        | 11 +-------
 gdb/symfile.c              |  1 -
 gdb/symmisc.c              |  4 +--
 gdb/symtab.c               | 23 +++++++++++++++--
 gdb/symtab.h               | 54 +++++++++++++++++++++++++++++++--------
 gdbsupport/Makefile.am     |  1 -
 gdbsupport/Makefile.in     |  5 +---
 gdbsupport/gdb-hashtab.cc  | 43 -------------------------------
 gdbsupport/gdb-hashtab.h   |  5 ----
 18 files changed, 116 insertions(+), 146 deletions(-)
---
base-commit: ecc5fed79101ba6af6ddc6e629e2d8d93c36065b
change-id: 20240605-hash-tab-no-obstack-91d863a9add4

Best regards,
-- 
Tom Tromey <tromey@adacore.com>