peirce@outpost.UUCP (Michael Peirce) (11/03/90)
In article <25483@dartvax.Dartmouth.EDU>, kurash@quimby.dartmouth.edu (Mark Valence) writes: > > I think what you are looking for is the Notification > Manager. The routines _NMInstall and _NMRemove are > both accessible during interrupt time, This is how it's _suppose_ to work. Unfortunately, NMInstall calls at interrupt level can cause some nice bombs with the current system software. I was asking DTS another question about the Notification Manager and in the reply they said something like: "and by the way, don't call NMInstall from interrupt level - it causes crashes sometimes. Sorry." That's it, no idea when it might be fixed or that it really seems to make the NM a lot less useful! I HAD been getting some random crashes when I used NMInstall at interrupt time, but I thought it was just some lurking bug. I guess it was a lurking bug, but not mine... So what is a well meaning programmer to do without NMINstall at interrupt time. The only thing I can think off it to patch System Task and have my interrupt code set flags to tell the patch to do a NMInstall... {This reminds me of the first couple of revs for the Time Manager that really didn't work either...) sigh :-( -- michael peirce -- Michael Peirce -- {apple,decwrl}!claris!outpost!peirce -- Peirce Software -- Suite 301, 719 Hibiscus Place -- Macintosh Programming -- San Jose, California 95117 -- and Consulting -- (408) 244-6554
zellers@prisoner.Eng.Sun.COM (If they build it) (11/04/90)
In article <b.D2QT9S@outpost.UUCP> kurash@quimby (Mark Valence) writes: >So what is a well meaning programmer to do without NMINstall at interrupt >time. The only thing I can think off it to patch System Task and >have my interrupt code set flags to tell the patch to do a NMInstall... I like to set flags which I read from a 'jGNEFilter' daisy chain - it's a nice place to patch, since it's called at a very stable time. -- -------------------------------------------------------------------------- Steve Zellers zellers@prisoner.sun.com "And that's why we have to have computers, because man, nobody's perfect."
kurash@carr.dartmouth.edu (Mark Valence) (11/06/90)
In article <2038@exodus.Eng.Sun.COM> zellers@prisoner.Eng.Sun.COM (If they build it) writes: >In article <b.D2QT9S@outpost.UUCP> kurash@quimby (Mark Valence) writes: >>So what is a well meaning programmer to do without NMINstall at interrupt >>time. The only thing I can think off it to patch System Task and >>have my interrupt code set flags to tell the patch to do a NMInstall... No, in fact I didn't write that. It was written by Michael Pierce, who pointed out that NMInstall is unstable. I must say, that changes things. I have always been hesitant about patching important traps (recall the discussion concerning patching ExitToShell and similar traps 'globally'), due to MultiFinder's intrusiveness. I might first write a driver that mimicked the Notification manager, install it at INIT time when my interrupt code is installed, and service these 'notification requests' at SystemTask (dNeedTime). >I like to set flags which I read from a 'jGNEFilter' daisy chain - it's >a nice place to patch, since it's called at a very stable time. >-- >-------------------------------------------------------------------------- >Steve Zellers zellers@prisoner.sun.com >"And that's why we have to have computers, because man, nobody's perfect." You said it, not me :) Mark.
d88-jwa@dront.nada.kth.se (Jon W{tte) (11/11/90)
In article <2038@exodus.Eng.Sun.COM> zellers@prisoner.Eng.Sun.COM (If they build it) writes: >I like to set flags which I read from a 'jGNEFilter' daisy chain - it's >a nice place to patch, since it's called at a very stable time. It's also a very, very discouraged way of doing things, since it's supposed to go away any day now... A patch to _OSEvent should be safer, I think. Or _SystemTask, as the original author said, since it's almost always called together with _OSEvent... h+ h+@nada.kth.se "Moof!(tm)"