Re: lynx-dev Lynx crash

[Klaus Weide]

On Thu, 16 Dec 1999, Larry W. Virden wrote:

> From: Klaus Weide <address@hidden>
> > On Sun, 12 Dec 1999, Larry W. Virden wrote:
> > > I was at mail.yahoo.com tonite and went to do an 'empty trash' operation
> > > and got a core dump.

> > Do you have "Links and form fields are numbered"?
> Yes

The reason I was asking: there are some scary comments in function
update_subsequent_anchors (in GridText.c), which only comes into
play with LINKS_AND_FIELDS_ARE_NUMBERED.

To quote part of the commentary:
     *   All that said, about the only time this will be a problem in
     *   *practice*, is when a page has near 1000 links or more (possibly
     *   after a TEXTAREA expansion), and has line crossing tag(s), and
     *   the tag numbers in a line crosser go from initially all 3 digit
     *   numbers, to some mix of 3 and 4 digits (or all 4 digits) as a
     *   result of the expansion process.  Oh, you also need a "clump" of
     *   anchors all on the same lines.
     *
     *   Yes, it *can* happen, but in real life, it probably won't be
     *   seen very much ...]

Some more scary comments in increment_tagged_htline, also only used
with LINKS_AND_FIELDS_ARE_NUMBERED.

I don't know whether this applies in your case, but...
It would be helpful to know whether crashes also occur without
LINKS_AND_FIELDS_ARE_NUMBERED.

> > "Inserting" how?
> Using ^V ^e in a textarea field to drop into my editor, then to read an
> amount of text (again - sorry I can't give you more specifics...)

Thanks.  (It could also have been INSERTFILE, or GROWTEXTAREA, or just
enter.)

> into the file, then left the editor, responded n to the wrapping question
> (I _never_ want to answer yes there - too bad I can't tell lynx not to bother
> asking me that question)

You reported a while ago that you even got the question when there
were no long lines to wrap.  Does *that* still happen?

> and then pressing the send button to 'submit the form'
> (I assume) to yahoo.

Something that might help track the problem down: make a copy
of the source of the page (the one with the textarea).  Try to
reproduce the crash with that copy.   (Don't actually submit -
At least I *suspect* that that's not needed, but going to some
other pages after the textarea manipulations may be needed to
trigger a crash.)

Maybe Kim has some better ideas.


   Klaus