[Ecls-list] bug in ecl_file_position_set

[Dustin Long]

Juan Jose Garcia-Ripoll wrote:
> On Dec 27, 2007 9:39 AM, Dustin Long <dlong at stevens.edu> wrote:
>   
>> Okay ignore that last patch! It seems that calling ecl_force_output on
>> an output stream in ecl_file_position_set shouldn't be done; it was
>> causing strange errors during shut down. This patch is better, though. I
>> tested it for a while and it didn't give me any problems.
>>     
>
> Hi Dustin,
>
> the first fix is the correct one. Obviously, one has to flush the
> buffers of an output stream before changing the file position, this is
> what the call to ecl_force_output() was doing. The second patch is
> wrong and I would like to know what problems you experienced during
> shut down because of the use of ecl_force_output(). Is it
> reproducible? On what platform?
>   
The problem is semi-reproducible, but I haven't tried too hard to figure 
out the steps. Basically I would play around in slime for a little bit 
(asdf load a collection of about 10 source files, run 3 or 4 functions) 
then quit, and get a crash. Strangely, when using only this patch, 
without the patch I submitted earlier (which changed how libraries where 
loaded), I would get heap corruption crashes much earlier, during 
runtime, but again have not figured out what functions would cause them. 
Maybe ecl_force_output() was getting called multiple times or something? 
I'll do some more work on it later to try and figure out more concrete 
steps. Platform is Windows XP.

Dustin