Lazarus caught AV, seems to be related to FpDebug
lazarus / 12-Debugger

----------------------------------------------------------------------
ALLIGATOR · 2025-11-11 08:55

Sorry, I can't repeat it. I tried, but it doesn't work...
But I took a screenshot and saved it, maybe it will give you some
ideas.

----------------------------------------------------------------------
ALLIGATOR · 2025-11-11 08:58

I also have the CPU-View package from Alexander Bagel installed.

I don't know if this could have any effect, because as far as I
understand, it is closely integrated.

----------------------------------------------------------------------
n7800 · 2025-11-11 10:05

It would be helpful if you provided the commit hash of your trunk.

----------------------------------------------------------------------
ALLIGATOR · 2025-11-11 10:25

A reasonable suggestion )

https://gitlab.com/freepascal.org/lazarus/lazarus/-
/commit/a486961c5f9f9472dcc3edd078cc146056e3ecb6

a486961c5f9f9472dcc3edd078cc146056e3ecb6

----------------------------------------------------------------------
Martin_fr · 2025-11-11 10:33

The sha1 wont make much of a diff, since there hadn't been many
changes in trunk.

Unfortunately its not much to go on.

It looks like (but could be something else too) some enumerator was
not freed (or not removed from the map).
But I can't find any potential enumerator use for this in the source.


Which version of FPC?

Because both, 3.2.2 and 3.3.1 have issues in the peephole optimizer
(at -O1 and up), so unless you compiled with -O- (no opts)...

3.2.3 seems to be ok.

----------------------------------------------------------------------
n7800 · 2025-11-11 10:46

Quote from: ALLIGATOR on November 11, 2025, 09:58:17 amI also have the
CPU-View package from Alexander Bagel installed.

I don't know if this could have any effect, because as far as I
understand, it is closely integrated.

This project? Yeah, it looks pretty serious.

----------------------------------------------------------------------
ALLIGATOR · 2025-11-11 10:59

Quote from: Martin_fr on November 11, 2025, 11:33:57 amUnfortunately
its not much to go on.   Sigh…

Quote from: Martin_frWhich version of FPC? FPC [main] of course! 

Quote from: Martin_frBecause both, 3.2.2 and 3.3.1 have issues in the
peephole optimizer (at -O1 and up), so unless you compiled with -O-
(no opts)...
-O- -O2 -gw3 -gl -gh -gt -Co -Cr -Ci -Sa
Different from the default, I experimented with this


Quote from: n7800 on November 11, 2025, 11:46:00 amThis project?Yes,
it is )

----------------------------------------------------------------------
Martin_fr · 2025-11-11 11:20

Quote from: ALLIGATOR on November 11, 2025, 11:59:28 amQuote from:
Martin_frWhich version of FPC? FPC [main] of course! 

Just in the last few weeks I run into 2 (maybe 1 and a half, if they
are related) peephole issues in 3.3.1. => I.e. FPC generates incorrect
asm code. And the effects of that, can of course manifest deferred, in
some semi unrelated code.

https://gitlab.com/freepascal.org/fpc/source/-/issues/41422
https://gitlab.com/freepascal.org/fpc/source/-/issues/41466