[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

bug#39962: 27.0.90; Crash in Emacs 27.0.90

From: Pip Cet
Subject: bug#39962: 27.0.90; Crash in Emacs 27.0.90
Date: Tue, 17 Mar 2020 13:54:05 +0000

On Tue, Mar 17, 2020 at 8:45 AM Pieter van Oostrum
<address@hidden> wrote:
> Pip Cet <address@hidden> writes:
> > On Tue, Mar 17, 2020 at 4:54 AM Pip Cet <address@hidden> wrote:
> >> m has not been set; it would have been set to 0x12590f570; can you
> >> print out the memory at that address? (x/32gx 0x12590f570)
> >
> > Actually, regardless of whether the pseudovector type is poisoned or
> > zero,

It's zero.

> > it would be good to dump that entire vector block (x/512gx
> > 0x12590f000)

> 0x12590f3f0:    0x0000000000000006      0x0000000000000000
> 0x12590f400:    0x000000000d4269c0      0x0000000000000000

I'm suspicious about this "symbol". Is 0xd4269c0 a valid lisp value?

> 0x12590f430:    0x0000000000000002      0x0000000000008df0
> 0x12590f440:    0x0000000005757688      0xa5a5a5a5a5a5a5a5

This vector is weird. The first entry is a symbol, but the second
entry looks invalid to me: all other symbols are aligned to 16-byte
boundaries, and I don't think 0x5757688 is even a valid pointer.

Can you check which symbol corresponds to 0x8df0 in your build? It
should be the one with 757 in its define line in globals.h

# define Qlibpng_version builtin_lisp_symbol (757)

> 0x12590f520:    0x4000000002002000      0x0000000200000003
> 0x12590f530:    0x000000011de53fc0      0xa5a5a5a5a5a5a5a5

A real-life bignum! Can you print x/2gx 0x11de53fc0 so we figure out
what its value is? (And what's creating bignums in your session?)

> 0x12590f540:    0x4000000003005000      0x000000015b8a3b30
> 0x12590f550:    0xa5a5a5a5a5a5a5a4      0x000000016082a820
> 0x12590f560:    0x0000000000014ba0      0x0000000000014ba0
> 0x12590f570:    0x0000000000000000      0x000000015b8a3b30

That's our corrupt word.

> 0x12590fbf0:    0x00000001079ffc00      0x0000000000000000

And that's the end of the vector block.

If you want to, you can try the attached patch and see whether it
produces anything poisoned rather than merely corrupt.

Thanks again!

Attachment: 0001-more-debugging.patch
Description: Text Data

reply via email to

[Prev in Thread] Current Thread [Next in Thread]