[squeak-dev] Croquet FFIPrim tests crash smalltalk on linux32

Nicolas Cellier nicolas.cellier.aka.nice at gmail.com
Tue May 12 21:10:07 UTC 2020


Hi all,
I confirm that FFI tests crashes a recent 32bits linux VM.
It segfaults in ffiPrintString which just printf the char * string passed...
The debug VM does not crash, which suggest that we still rely on some
undefined behavior.
I did retry my one shot gun: recompile with clang
(see https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/445)

   $ CC=clang ./mvm

and the FFI tests then passed in optimised version too...
I suggest that we use that gun on the CI until we find the problem.

Le mar. 12 mai 2020 à 00:12, Vanessa Freudenberg <vanessa at codefrau.net> a
écrit :

> Not sure about those crashes, but does Nikolay's stuff work?
>
> https://blog.krestianstvo.org/en/open-croquet-for-squeak-6/
>
> - Vanessa -
>
> On Mon, May 11, 2020 at 12:40 PM gettimothy via Squeak-dev <
> squeak-dev at lists.squeakfoundation.org> wrote:
>
>> Hi folks,
>>
>> Since the Croque works on the old image, I decided to try getting
>> CroquetGL working on Squeak5.3/6.0alpha
>>
>> In both cases, running the FFI tests crashes the image with stuff like
>> this (this is what I copy paste from single xterm screen, let me know if
>> you need redirect to file)
>>
>>
>> utcMicrosecondClock
>> //
>> **StackOverflow**
>> vmParameterAt:
>> vmParameterAt:
>> utcMicrosecondClock
>> getSystemAttribute:
>> getSystemAttribute:
>> basicNew
>> basicNew:
>> stackp:
>> resume
>> value
>> wait
>> signal
>> utcMicrosecondClock
>> +
>> <
>> <=
>> >=
>> +
>> <
>> primSignal:atUTCMicroseconds:
>> wait
>> signal
>> wait
>> signal
>> suspend
>> getSystemAttribute:
>> getSystemAttribute:
>> basicNew:
>> stackp:
>> resume
>> value
>> wait
>> signal
>> utcMicrosecondClock
>> +
>> <
>> <=
>> >=
>> +
>> <
>> primSignal:atUTCMicroseconds:
>> wait
>> signal
>> wait
>> perform:
>> hashWithInitialHash:
>> compare:with:collated:
>> at:
>> headerWord
>> hashWithInitialHash:
>> compare:with:collated:
>> invokeWithArguments:
>>
>> stack page bytes 4096 available headroom 2788 minimum unused headroom 1704
>>
>>         (Segmentation fault)
>> Aborted
>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20200512/403726ce/attachment.html>


More information about the Squeak-dev mailing list