[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: TCP problems only on *some* macintoshes...?

At 13:09 10/3/94 -0600, Joey Gray wrote:
>At  7:06 AM 10/3/94 -0500, you wrote:
>>There is a problem with MacTCP.lisp on newer machines which I haven't
>>fully investigated, but for which the workaround is to turn off the ASYNC
>>bit in the #_Control trap in the TCP-CONTROL function.  In my case a IIfx
>>which simply inherited a newer ethernet board would intermittently hang
>>or crash when using TCP from MCL (but not from other applications). See
>>if this helps.  Let me know if it does, I'll try to push up the priority of
>>investigating what's really going wrong...
>i looked in my copy of MacTCP.lisp, and only found two instance of the
>#_Control trap being called (in %tcp-control and tcp-active-open). since
>the one in %tcp-control was the only one that mentioned async, i changed
>the following line in that function:
>          (when (eql (setq err (#_control :async pb)) 0)
>          (when (eql (setq err (#_control pb)) 0)
>is that what you meant?
>so i made a new version of our application with this change, and have sent
>it off to the two people having this problem, and i'll soon know whether
>this fixes our problem, or not. thanks a ton!

Yes, this was indeed the change I was suggesting.  It is a bandaid, I've still
got it on my queue to figure out what's really going wrong.

>>>ps - how can i get a list of all of those patches that bill st. clair
>>>(that's the guy, right?) keeps talking about, and then how do i get these
>>Well, there's a set of patches that was a candidate for being released, but
>>I discovered that the MCL system wouldn't correctly compile itself with
>>all the patches loaded and haven't yet had the time to identify and either
>>withdraw or fix the culprit(s).  When I do, we'll release that set.
>>Since that set was put together, of course, both Bill and I have been
>>accumulating more patches as bugs get reported and fixed.  Unfortunately,
>>creating a patch to fix a single case that a customer reported is different
>>from ensuring that it works in all possible cases and therefore is worthy of
>>widespread release (as the experience with the last set which we THOUGHT
>>was ready for release shows), so I'm hesitant to publicize the list.
>>However, if you have a bug that's bothering you (or even suspect a possible
>> bug in some area), by all means send mail asking about it.
>actually, i wasn't talking about the above problem at all. on various
>occasions, i've seen bill say something about 'send mail and ask for the
>make-lisp-faster patch' and i was just curious about that. i wanted to get
>a list of those patches (and short descriptions of each) and to find out
>how to get this information. is there a listserv? i looked at the cambridge
>ftp site, but didn't find much, and i'm pretty sure that i've seen bill
>announce more fixes than were available there. sorry i wasn't too clear...

Yes, those were the patches I was talking about too.  All the developers make
such patches, and keep a local stash of them, and send them out in reponse
to bug-reports.  However, until we do a "collection" of bug fixes, there isn't
any centralized list of what ones exist and what each one fixes nor is there
any systematic testing of a particular patch's effect on the rest of the system.
Often the patch doesn't quite fix things, or introduces other bugs, or uncovers
other bugs which have been hidden.  We don't want to just make them available
wholesale, because that would be likely to cause our customers more grief
than the status quo.

We try to collect patches, look at them and test them thoroughly enough to
ensure that they don't break building and running the Lisp system a couple of
times a year, but have recently been very much swamped under the pressure
of other things.  But it is still on the list.

>thanks again! i'll let you know if your fix solves our problems!