[0/4] reftable: fix realloc error handling

Message

René Scharfe Dec. 25, 2024, 6:33 p.m. UTC

The current handling of reallocation errors leaks the original
allocation in most cases and corrupts the capacity variable.  Fix
that in REFTABLE_ALLOC_GROW and by providing a new macro
REFTABLE_ALLOC_GROW_OR_NULL -- solve this somewhat tricky issue
centrally, with minimal impact to calling code.

And the last two patches add error handling to the remaining
places that still lack it.

  reftable: avoid leaks on realloc error
  reftable: fix allocation count on realloc error
  reftable: handle realloc error in parse_names()
  t-reftable-merged: handle realloc errors

 reftable/basics.c                | 14 ++++-------
 reftable/basics.h                | 41 +++++++++++++++++++++++++-------
 reftable/block.c                 | 10 ++++----
 reftable/pq.c                    |  2 +-
 reftable/record.c                | 12 +++++-----
 reftable/stack.c                 |  8 ++++---
 reftable/writer.c                |  5 ++--
 t/unit-tests/t-reftable-merged.c |  4 ++--
 8 files changed, 60 insertions(+), 36 deletions(-)

--
2.47.1

Comments

Patrick Steinhardt Dec. 27, 2024, 10:34 a.m. UTC | #1

On Wed, Dec 25, 2024 at 07:33:07PM +0100, René Scharfe wrote:
> The current handling of reallocation errors leaks the original
> allocation in most cases and corrupts the capacity variable.  Fix
> that in REFTABLE_ALLOC_GROW and by providing a new macro
> REFTABLE_ALLOC_GROW_OR_NULL -- solve this somewhat tricky issue
> centrally, with minimal impact to calling code.
> 
> And the last two patches add error handling to the remaining
> places that still lack it.

Thanks a lot for working on this!

Patrick

Junio C Hamano Dec. 27, 2024, 4:02 p.m. UTC | #2

Patrick Steinhardt <[e-mail]> writes:

> On Wed, Dec 25, 2024 at 07:33:07PM +0100, René Scharfe wrote:
>> The current handling of reallocation errors leaks the original
>> allocation in most cases and corrupts the capacity variable.  Fix
>> that in REFTABLE_ALLOC_GROW and by providing a new macro
>> REFTABLE_ALLOC_GROW_OR_NULL -- solve this somewhat tricky issue
>> centrally, with minimal impact to calling code.
>> 
>> And the last two patches add error handling to the remaining
>> places that still lack it.
>
> Thanks a lot for working on this!

And thank you for reviewing.