Re: Strange Segfault with V 0.9.3 an MinGW Cross Compile

[aris <aris@xxxxxxxxxx>]

Hi,I looked at the generated assembly and don't see anything weird. As you said it might be a strange mismatch between compile time and runtime libc.A stacktrace would help.ArisEnvoyé depuis mon smartphone Samsung Galaxy.
-------- Message d'origine --------De : "Skalák, Zdeněk" <zskalak@xxxxxxxxxxxx> Date : 15/01/20  10:03  (GMT+01:00) À : libssh@xxxxxxxxxx Objet : Re: Strange Segfault with V 0.9.3 an MinGW Cross Compile Hello,On Wed, Jan 15, 2020 at 9:48 AM <g4-lisz@xxxxxxxxxxxx> wrote:On 15.01.20 08:27, Andreas Schneider wrote:

> On Wednesday, 15 January 2020 01:22:13 CET g4-lisz@xxxxxxxxxxxx wrote:
>> I cross compiled V 0.9.3 with MinGW on Ubuntu. The compilation worked
>> flawlessly.
>>
>> But when I run my code linked to the resulting libssh.dll, it crashes
>> immediately when I set verbosity to > 2.
>>
>> I could track this down to vsnprintf() which is called through line 865
>> in channels.c:
>>
>>   SSH_LOG(SSH_LOG_PACKET,
>>           "placing %zu bytes into channel buffer (%s)",
>>           len,
>>           s_stderr ? "stderr" : "stdout");
> Try s_stderr -> is_stderr

Hi Andreas,

Haha that would be a silly issue, when the compiler would not complain
about this...

Sorry, I somehow lost the 'i' when I pasted the code into the mail.I can imagine two things which could went wrong:a) libstdc (or whomever implements the vsnprintf()) doesn't know the `%zu` format, so this one is skipped and the `len` argument is used for the (%s) formatb) 32/64 bit mismatch - the format (count of bytes) of the `len` on the stack doesn't correspond to the format (count of bytes) of the `%zu` specifier, so for the second (%s) format, some `grabage from the stack` is usedBoth these two scenarios can lead to mysterious crashes ....RegardsZdenek OGAR Skalak

Cheers,
Till


-- Ing. Zdeněk OGAR SkalákMonet+ a.s.            <http://www.monetplus.cz>Za Dvorem 505, 763 14 Zlín - Štípa, CZTel: +420 / 577 110 411,  Fax: +420 / 577 914 557