Re: [Audacity-devel] [DUMP+FREEBSD] Closing audacity 2.3.3_1 leads to core dump

[Paul L. <pau...@gm...>]

I'm looking more closely.  I think there is more than one thing going on
here.

I think there really is a bug in the event handlers, which may happen or
not depending on the ordering of loading of Scrubber and ScrubUI.  I never
saw it on Mac, and quite possibly it happened only on certain builds of
certain flavors of Linux, and then did start happening on Windows with
CMake.

The proposed fix isn't the right one.  It eliminates the pushing of the
handler.  No.  The pushing and popping should instead be the responsibility
of one class, not divided between two classes.  I will push the fix I think
is right.

However I'm not yet convinced that this explains the very recent bug report
2332.  Something else is happening.

I'm trying to bisect it.

PRL


On Fri, Feb 21, 2020 at 1:49 PM Leland Lucius <ll...@ho...> wrote:

> Correct, it is not recent.  If this is truly the same issue, It was first
> reported by Jack L. on this list back in November:
>
>
>
> https://sourceforge.net/p/audacity/mailman/message/36859764/
>
>
>
>
>
> *From:* Paul Licameli <pau...@gm...>
> *Sent:* Friday, February 21, 2020 11:53 AM
> *To:* Hans Petter Selasky <hp...@se...>
> *Cc:* Paul Licameli <pau...@au...>; Audacity-Devel list
> <aud...@li...>
> *Subject:* Re: [Audacity-devel] [DUMP+FREEBSD] Closing audacity 2.3.3_1
> leads to core dump
>
>
>
> Sorry I got overwhelmed and overlooked this thread.  Scream at me in the
> subject line next time if you really need my attention.
>
>
>
> Leland merged this patch, but I reverted it.
>
>
>
> I don't believe this explanation for the recent regression.  You blamed
> the problem on commit a3b434be1cd42ff7c99821de4ea52c0945cb0baa, but that is
> not recent.
>
>
>
> There may be some other explanation of unbalanced push and pop of event
> handlers.  Maybe some other faulty code makes an unbalanced pop first, and
> the destructor of Scrubber is innocent, but appears in the stack trace.
>
>
>
> I'll be more convinced if you figure out where the crash happens after
> bisecting the history of commits only in version 2.4.0.
>
>
>
> PRL
>
>
>
>
>
>
>
> On Thu, Feb 13, 2020 at 3:22 AM Hans Petter Selasky <hp...@se...>
> wrote:
>
> On 2020-02-13 07:37, Jack L. wrote:
> > *** BUG ***
> > In void pixman_region32_init_rect(region_type_t *, int, int, unsigned
> > int, unsigned int): Invalid rectangle passed
> > Set a breakpoint on '_pixman_log_error' to debug
>
> I traced this one to the width of a window being -4U . I didn't
> investigate further.
>
> --HPS
>
>