Re: pvm_execute_or_initialize SEGV with reproducer

[David Faust]

On 4/22/21 1:30 PM, Jose E. Marchesi wrote:
> 
>> Hi David.
>>
>> Thanks for reporting.
>>
>>> The same also occurs with none of the --disables, they serve to simplify
>>> the backtrace.
>>>
>>> (poke) .file ~/playpen/btf/poke-fault/poke-fault.o
>>> (poke) load elf
>>> (poke) load "btf-dump.pk"  # note, this also loads btf.pk
>>> (poke) var elf = Elf64_File @ 0#B
>>> (poke) var shdr = elf.get_sections_by_name (".BTF") [0]
>>> (poke) var btf = BTF_Section @ shdr.sh_offset
>>
>> I took a quick look, and it seems that the exceptions stack is being
>> overflown.  This is probably a code generator bug: somehow we are not
>> executing enough `pope' instructions for all the `pushe'.
>>
>> As far as I can see, we are starting to "leak" elements in the
>> exceptions stack in:
>>
>>   (poke) var btf = BTF_Section @ shdr.sh_offset
>>
>> As soon as I have a little time I will look at all the places where the
>> compiler is generating pushe/pope pairs and see what is going on..
>> unless someone beats me to it :)
> 
> 
> Ok this is now fixed by the commit below, which I have pushed to both
> master and maint/poke-1.
> 
> It should fix your issue with btf-dump.

Yep, looks like this fixes the issue (and wow, that was quick!)

Thanks!

> 
> 
>     pkl-gen: avoid leaking slots in the exceptions stack
>     
>     2021-04-22  Jose E. Marchesi  <jemarch@gnu.org>
>     
>             * libpoke/pkl-gen.pks (struct_mapper): Avoid "leaking" slots in
>             the exceptions stack.
>             (struct_field_mapper): Likewise.
>