Re: [PATCH] Re: freeze for sbcl-2.6.2

This AI is clever, but not clever enough to see that it's already
loaded into temp in a previous instruction.

On Wed, Feb 25, 2026 at 4:02 AM Andreas Franke via Sbcl-devel
<[email protected]> wrote:
>
>
>
> It's only :allocation-size-histogram that's the problem. :cons-profiling is not factor.
> My focus is on the per-thread allocation-tracking (not using the full histograms).
> Asking the AI to bisect leads it to suggest the attached fix. (I really didn't add anything meaningful, just upgraded the "co-author" to "author".)
> On my machine, both tests that fail on current master=sbcl-2.6.1-221-g5e6717f7e appear to succeed with this patch applied.
>
> I could make them mutually exclusive selections in src/cold/shared, but let me guess: you use both. Is making that restriction a problem?
> I think it's possible to derive the histogram from the raw allocations.
>
> On Sun, Feb 22, 2026 at 9:07 PM Andreas Franke via Sbcl-devel <[email protected]> wrote:
>>
>> Building with both :allocation-size-histogram and :cons-profiling on linux x86-64,
>> I'm seeing unexpected test failures on current master (sbcl-2.6.1-205-gb46b7221f),
>> both pointing to the same issue:
>
> _______________________________________________
> Sbcl-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/sbcl-devel


_______________________________________________
Sbcl-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sbcl-devel