Re: lynx-dev dev11 core dump in LYEdit.c under DJGPP

[dickey]

> 
> >> 
> >> The new problem appears a couple of dev versions ago: 
> >> when I look localfile with DJGPP port and press 'e' to edit the file 
> >> I got a crash immediately. 
> >> 
> >> gdb shows a crash from extra free() in 
> >> edit_current_file() at LYEdit.c:129 (quoted below): 
>  
> > thanks (will check/fix) 
>  
> >> #ifdef __DJGPP__ 
> >>         sprintf(command, "%s %s", editor, HTDOS_name(filename)); 
> >> #else 
> >>         sprintf(command, "%s \"%s\"", editor, HTSYS_name(filename)); 
> >> #endif /* __DJGPP__ */ 
> >> #endif /* VMS */ 
> >>     CTRACE(tfp, "LYEdit: %s\n", command); 
> >>     CTRACE_SLEEP(MessageSecs); 
> >> #ifndef __EMX__ 
> >> ---->    FREE(filename); 
> >> #endif 
> >> 
> >>     /* 
> >>      *  Invoke the editor. - FM 
> >> 
> >> 
> >> I change the condition to 
> >> #if !defined(__EMX__) && !defined(__DJGPP__) 
> >> and it works OK but I do not know the origin of the problem 
> >> so expect problems elsewhere. 
>  
> > well, it doesn't seem to be directly related to the HTSprintf changes. 
> > Perhaps I moved an ifdef. 
>  
> No. I checked against 2.8.1rel: 
> LYEdit.c, HTSYS_name(), HTDOS_name() was not changed but the problem come. 
> Probably certain memory problems was hidden until you change something. 

with the last few patches, I've replaced some fixed buffers with ones
that are allocated.  I've _looked_ (but can miss) places where the
calling code extends the buffer by simplying concatenating new characters
onto it.  Perhaps that is the problem.
  
> In fact, HTSYS_name() equivalent to HTDOS_name() for __EMX__ and DOSPATH, 
> but HTDOS_name() do the odd thing - it returns a pointer to initial argument 
> which is modified.  This is not intended for HTVMS_name() so we got 
> HTDOS_name() argument converted several times in edit_current_file() 
> and possibly other places. Should we fix HTDOS_name ? 

probably a good thing to fix.  I was looking at this (or something
similar) the other night, and it was returning a pointer to a static (1024)
buffer.  I wanted to replace the buffer with an allocated one, but the
calling code was a little messy - some places it seems to assume that
the buffer is longer than the string.  So I put it off for another day.
  
> Trace have the folowing entries: 
>  
> HTParse: result:/c:/lynx/dist28/lynx2-8-/src/chrtrans/cp1250_u.h 
> HTDOS_name changed `c:/lynx/dist28/lynx2-8-/src/chrtrans/cp1250_u.h' to 
> `c:\lynx\dist28\lynx2-8-\src\chrtrans\cp1250_u.h' 
> HTDOS_name changed `c:\lynx\dist28\lynx2-8-\src\chrtrans\cp1250_u.h' to 
> `c:\lynx\dist28\lynx2-8-\src\chrtrans\cp1250_u.h' 
>  
>  
> > -- 
> > Thomas E. Dickey 
>  
>  


-- 
Thomas E. Dickey
address@hidden
http://www.clark.net/pub/dickey