Re: LYNX-DEV Re: Lynx updates

[Klaus Weide]

On Sun, 10 Nov 1996, Foteos Macrides wrote:

>[ about the DIRED patches ] 
>       Note that I again had to remove the code for setting the Lynx
> User-Agent string as an environment variable.  It had been included
> in the patch, without comment that it had been included.  Please be
> more careful about that, OK?

Please accept my apologies for that,  it definitely shouldn't have
happened. 

>[ Klaus Weide <address@hidden> wrote:]
> 
> >Now, is someody gonna try those damn patches and report on it, or do I
> >have to go on and on in long-winded messages [...]
> 
>       Please do your best not to be long-winded, but we'll try to
> cope with it if there's no alternative. :) :) :)

Well it seems to have gotten me somewhere :)

I wish Rob could be a *little* more long-winded about the changes he
is making;  that way he probably could get more testers for his latest
colour style changes.   
[Are you listening Rob? :)]

   ---
And now something completely different:

We are trying hard[*] in the Lynx code to clean up all memory that
is alloc'd - in an atexit() function if not earlier.  (I don't know
for what systems' benefit we have to be so diligent..)  My observation
for tha day is that, given the current (and future) architecture of
the Library code, cleaning up all memory cannot fully succeed if any
HTStream objects are active at the time of the atexit() call.  The only
way around that would be for all HTStream objects to register themselves
somehow somewhere on creation so that the atexit() could find them.

If the exit() results from a Fatal Error, that's probably acceptable -
things are messed up anyway.  But there is code that calls exit()
explicitly from within the HTStream pipe, in HTFWriter.c (I haven't checked
other places).  I am wondering whether it is really necessary to have
HTFWriter_free (this for dump_output_immediately only) exit() at that
point, rather than first returning "safely" to the MainLoop or at least
the HTLoad().  I haven't checked more closely whether this ever actually
creates a memory leak in the current Lynx code - it probably will not
as long as the HTFWriter is the only HTStream in the pipe.  But I ran
into weird effects when playing with this kind of thing in connection 
with the libwww5.  Any comments?

Yes, I should try the LYleaks stuff, it looks neat, someday I may get
around to it..

  Klaus

[*] Fote is trying hard, and I am trying to learn. [**]
[**] Do footnotes make it appear less long-winded? :)

;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;