mbox series

[0/5] Fix race in DWARF reader, 2nd approach

Message ID 20240416-dwarf-race-relocate-2-v1-0-1fc912e95e87@tromey.com
Headers
Series Fix race in DWARF reader, 2nd approach |

Message

Tom Tromey April 16, 2024, 5:05 p.m. UTC 
  This series is a different approach to fixing the race pointed out in
PR gdb/31261.  The problem there is that the indexer can be running at
the same time that an objfile is relocated, resulting in races in
dwarf2_per_objfile::adjust.

The previous series to fix this problem is here:

https://inbox.sourceware.org/gdb-patches/20240217-dwarf-race-relocate-v1-0-d3d2d908c1e8@tromey.com/

This series is less invasive and works by justifying the eventual
removal of the 'adjust' method.

Regression tested on x86-64 Fedora 38.

I also tested this in a small way on a MIPS executable.  I don't have
access to a machine where I can truly test MIPS16 in its entirety, but
what I did is make an executable where one function was marked
__attribute__((mips16)).  Then, I tested that the symbol side of gdb
still correctly applies the address transform in "break" and
"disassemble" commands with this series applied.

---
Tom Tromey (5):
      Remove call to dwarf2_per_objfile::adjust from ranges readers
      Remove more calls to dwarf2_per_objfile::adjust
      Remove call to dwarf2_per_objfile::adjust from read_call_site_scope
      Remove call to dwarf2_per_objfile::adjust from read_attribute_value
      Remove dwarf2_per_objfile::adjust

 gdb/dwarf2/aranges.c        |  2 --
 gdb/dwarf2/read-gdb-index.c |  2 --
 gdb/dwarf2/read.c           | 45 ++++++++++-----------------------------------
 gdb/dwarf2/read.h           |  4 ----
 4 files changed, 10 insertions(+), 43 deletions(-)
---
base-commit: 12f5356130c2cda10e2589e74a8716563050dccb
change-id: 20240416-dwarf-race-relocate-2-621894b33715

Best regards,