Re: lynx-dev libreadline

[brian j pardy]

On Sweetmorn, the 54th of The Aftermath, 3165, Klaus Weide wrote:
> On Sun, 12 Dec 1999, brian j pardy wrote:
> 
> > Yeah, I've still been planning on taking a look.  My home machine was
> > having a lot of problems (motherboard going out), so I had been trying
> > to do as little as possible on it.  Now that I bought a new box, I'm
> > slowly building everything back up to spec and will very soon be able
> > to get back to it.
> > 
> > Re-reading the posts from that thread, I think your (second) proposal
> > is the one that I find most attractive.  
> 
> I'm not sure it's the better one; it's more flexible, but requires a
> bit more typing.  Well for batch-like use, that should be ok.  Glad
> to leave the decision to you (or others).

I don't want to make the decision, but since nobody else seems to
want to code it... :)

I need to do this right now -- new NetHack out just in time to eat up
my dev time.

I'll be making it a lynx.cfg and .lynxrc option as well.  Any
objections?  I can see a few from it being "only" useful in batch
mode, but I like the ideas about implementing a read-only cookie file
for multiple users.

> > > [...] The only thing I find ugly about it is
> > > that "/dev/null" is an OS-specific convention.
> > 
> > Is there anything in Lynx to mimic output to /dev/null for non-UNIX?
> > 
> > I'd rather not do an if(!unix && writefile != /dev/null) writecookies; 
> 
> You don't really need to, just check for "/dev/null" (if we say that
> that should have Unix-ish /dev/null meaning on other systems, too).
> If the user prefers to use a non-Unix system's native equivalent -
> like "NUL:" (or "NUL"?) for DOS, etc., and "NL:" for VMS - it should
> *automatically* have the same effect, the system will take care of
> it.

Do we give a Unix-ish meaning to /dev/null in other places?  I think
it should have that meaning, but I'd want to be consistent.

> > sort of thing -- has anyone solved this problem before?
> 
> The is some VMS-specific stuff in some files in src/, grep for "nl:".
> But again, if you stick with just opening the file for writing, write,
> close, as it is now, then no special recognizing of those names should
> be necessary.

*nod*

> Actually, what I'd like to see (and maybe one day implement, but not
> now - please do so if you have the time) is a sefaer way for writing
> files.  At least for "important" user files: .lynxrc and cookie files.
> Lynx should creat a new temp file for writing, and only replace the
> original file if writing was successful.  Somewhat like for bookmark
> files when removing (but that code may be more complicated than
> necessary).  In *that* case, recognizing OS-specific "NUL:" and "NL:"
> names would probably be necessary.

I'll take a look at the bookmark code, see what I can come up with.
Perhaps this is something that should be done in conjunction with
multiple processes writing to the file -- while I can't see that
happening too often with .lynxrc, I suppose it COULD.

-- 
My theology, briefly, is that the universe was dictated but not signed.
                -- Christopher Morley