mcmahan@netcom.UUCP (Dave Mc Mahan) (10/14/90)
I was just cruising through some software today, and noticed that many, many,
programs leave the dos.library open after exiting. In fact, it seems to be
a majority in the (limited) sample of stuff I tested, mostly PD. Upon looking
further, I find that my Manx 'C' 3.4 compiler seems to generate programs that
open the dos.library, but never seems to close it upon exit. If most folks
where using the Manx compiler (or if Lattice did it also) I can see that this
would indeed leave open copies of dos.library around upon exit. One would
assume that the compiler startup code would Close() anything that was Open()ed.
Is this a bug, or am I missing something? I am in the process of ripping
apart the actual startup code that Manx links in, but my disassembler seems
to find no place where the library is closed. Why is this?
I guess the other option is that Xoper is lying to me and there really aren't
multiple opennings of dos.library around (not very likely).
-davecedman@lynx.ps.uci.edu (Carl Edman) (10/14/90)
In article <14721@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes:
I was just cruising through some software today, and noticed that many,many,
programs leave the dos.library open after exiting. In fact, it seems to be
a majority in the (limited) sample of stuff I tested, mostly PD.Upon looking
further,I find that my Manx 'C' 3.4 compiler seems to generate programs that
open the dos.library, but never seems to close it upon exit. If most folks
where using the Manx compiler (or if Lattice did it also)I can see that this
would indeed leave open copies of dos.library around upon exit. One would
assume that the compiler startup code would Close() anything that was
Open()ed.
Is this a bug, or am I missing something? I am in the process of ripping
apart the actual startup code that Manx links in, but my disassembler seems
to find no place where the library is closed. Why is this?
I guess the other option is that Xoper is lying to me and there really aren't
multiple opennings of dos.library around (not very likely).
First, all my knowledge is about DOS 1.3. I haven't programmed on the amiga
too much recently, so if anything drastic has changed in 2.0, disregard
this post.
Yes, I have noticed that many programms do this. It is not very nice and
very clean, but in the case of "dos.library" it doesn't really matter.
There is only one copy of "dos.library" around anyway and that copy
is not in RAM but in WOM (yes, for you people out there who use these
new-fangled (:-) 500s and 2000s that is ROM). Even if you could
expunge it , it wouldn't be very likely that you would ever with a
running system get in the situation where no task anylonger needs
the dos.library. So never mind about that. It is only a matter of
tidyness (or neuroticism if you prefer).
Ever noticed that no programm ever even opens much less closes
"exec.library" ? Now that is a serious problem which you should
investigate (Lots of :-) in this paragraph.)
Carl Edman
PS: I always opened and closed not only "dos.library" but even "exec.library".
(Yes, I am a neurotic :-).
Theorectial Physicist,N.:A physicist whose | Send mail
existence is postulated, to make the numbers | to
balance but who is never actually observed | cedman@golem.ps.uci.edu
in the laboratory. | edmanc@uciph0.ps.uci.edumcmahan@netcom.UUCP (Dave Mc Mahan) (10/15/90)
In a previous article, cedman@lynx.ps.uci.edu (Carl Edman) writes: >In article <14721@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes: > I was just cruising through some software today, and noticed that many,many, > programs leave the dos.library open after exiting. In fact, it seems to be > a majority in the (limited) sample of stuff I tested, mostly PD. > >Yes, I have noticed that many programms do this. It is not very nice and >very clean, but in the case of "dos.library" it doesn't really matter. >There is only one copy of "dos.library" around anyway and that copy >is not in RAM but in WOM (yes, for you people out there who use these >new-fangled (:-) 500s and 2000s that is ROM). Even if you could >expunge it , it wouldn't be very likely that you would ever with a >running system get in the situation where no task anylonger needs >the dos.library. So never mind about that. It is only a matter of >tidyness (or neuroticism if you prefer). I realize that it is generally no big deal, since one would never want to expunge the dos.library, but it seems to indicate a possible problem. For those that want to leave the amiga on for weeks or months at a time and run lots of programs (we are talking about thousands if not billions of program executions) that leave dos.library open, wouldn't the variable eventually overflow and possibly cause problems. I'm messy in the rest of my life, but I try to make my programs thoughtful and tidy. Leaving a library open is not tidy. Who knows, it might cause bad effects in AmigaDos 5.3 or higher. It's just a good idea. >Ever noticed that no programm ever even opens much less closes >"exec.library" ? Now that is a serious problem which you should >investigate (Lots of :-) in this paragraph.) I guess it would be kind of tough to close, but it could be done. Could be the basis for a real neat virus if you expunged the exec.library. :-) > > Carl Edman > >PS: I always opened and closed not only "dos.library" but even "exec.library". >(Yes, I am a neurotic :-). Good to see. -dave
cedman@lynx.ps.uci.edu (Carl Edman) (10/15/90)
In article <14766@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes: In a previous article, cedman@lynx.ps.uci.edu (Carl Edman) writes: >In article <14721@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes: > I was just cruising through some software today, and noticed that many,many, > programs leave the dos.library open after exiting. In fact, it seems to be > a majority in the (limited) sample of stuff I tested, mostly PD. > >Yes, I have noticed that many programms do this. It is not very nice and >very clean, but in the case of "dos.library" it doesn't really matter. >There is only one copy of "dos.library" around anyway and that copy >is not in RAM but in WOM (yes, for you people out there who use these >new-fangled (:-) 500s and 2000s that is ROM). Even if you could >expunge it , it wouldn't be very likely that you would ever with a >running system get in the situation where no task anylonger needs >the dos.library. So never mind about that. It is only a matter of >tidyness (or neuroticism if you prefer). I realize that it is generally no big deal, since one would never want to expunge the dos.library, but it seems to indicate a possible problem. For those that want to leave the amiga on for weeks or months at a time and run lots of programs (we are talking about thousands if not billions of program executions) that leave dos.library open, wouldn't the variable eventually overflow and possibly cause problems. I'm messy in the rest of my life, but I try to make my programs thoughtful and tidy. Leaving a library open is not tidy. Who knows, it might cause bad effects in AmigaDos 5.3 or higher. It's just a good idea. I just checked the source to exec.library. Two points: a) The open_cnt really only is a UWORD i.e. its maximal value is 65535, a value which could possibly be reached in month long execution. b) OpenLibrary() doesn't itself do anything about open_cnt but expects the library open routine to take care of that. So it is possible that a badly written library, in conjunction with badly written program, in conjunction with veyr long running could lead to problems like this: 1. Programm BadProgram gets loaded and opens BadLibrary , but as it is a BadProgram it doesn't close it again. open_cnt stays 1 to large. 2. GoodUser leaves the computer running for months during which time BadProgramm gets loaded 65536 times. 3. As BadLibrary doesn't check for overflow it increases the open_cnt to 0 from 65535 when it is called by BadProgram again. 4. While BadProgramm still runs there is a memory crunch. All libraries are checked for expunging and BadLibrary has an open_cnt of 0 and is expunged 5. BadProgram accesses BadLibrary which is gone and ** CRASH ** ! All of this could be prevented if Exec OpenLibrary() took care of of open_cnt and other such details and checked for overflow. Why doesn't it ? Last irrelevant question: Does dos.library check for overflow ? (Not that it matters. It can't be expunged). Carl Edman Theorectial Physicist,N.:A physicist whose | Send mail existence is postulated, to make the numbers | to balance but who is never actually observed | cedman@golem.ps.uci.edu in the laboratory. | edmanc@uciph0.ps.uci.edu
kris@kriski.toppoint.de (Kristian Koehntopp) (10/16/90)
In article <14721@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes: > I was just cruising through some software today, and noticed that many,many, > programs leave the dos.library open after exiting. In fact, it seems to be > a majority in the (limited) sample of stuff I tested, mostly PD. This is due to an error in the startup-code of Manx Aztec C Compiler, which was present to at least version 3.6. This startup-code opened the dos.library but did not close it automatically. I always added an appropiate "CloseLibrary" to my programs. Kristian I also have to pay for incoming mail. Do not reply or keep short! Kristian Koehntopp, Harmsstrasse 98, 2300 Kiel, +49 431 676689, kris@tpki.UUCP Eine Nachricht dieses Absenders enthaelt 235 Bytes ASCII und 2% Header (mind.) nach DIN. Der Bundesinformationsminister: Net-News gefaehrdet Ihre Freizeit!
rhialto@cs.kun.nl (Olaf Seibert) (10/17/90)
In article <CEDMAN.90Oct14141233@lynx.ps.uci.edu> cedman@lynx.ps.uci.edu (Carl Edman) writes: > 4. While BadProgramm still runs there is a memory crunch. > All libraries are checked for expunging and BadLibrary has > an open_cnt of 0 and is expunged But BadLibrary needs to expunge _itself_, so if it is really not well behaved it doesn't even do that. > Carl Edman -- Olaf 'Rhialto' Seibert rhialto@cs.kun.nl How can you be so stupid if you're identical to me? -Robert Silverberg
plav@cup.portal.com (Rick M Plavnicky) (10/18/90)
In a recent article kris@kriski.toppoint.de (Kristian Koehntopp) writes: }In article <14721@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes: }> I was just cruising through some software today, and noticed that many,man programs leave the dos.library open after exiting. In fact, it seems to be }> a majority in the (limited) sample of stuff I tested, mostly PD. } }This is due to an error in the startup-code of Manx Aztec C Compiler, which }was present to at least version 3.6. This startup-code opened the }dos.library but did not close it automatically. I always added an appropiate }"CloseLibrary" to my programs. } }Kristian I just looked at this in Manx 5.0d, and apparently dos.library is now closed properly at exit time. /* Rick Plavnicky ...!sun!cup.portal.com!plav -or- plav@cup.portal.com */
mcmahan@netcom.UUCP (Dave Mc Mahan) (10/18/90)
In a previous article, kris@kriski.toppoint.de (Kristian Koehntopp) writes: >In article <14721@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes: >> I was just cruising through some software today, and noticed that many >> programs leave the dos.library open after exiting. It seems to be >> a majority in the (limited) sample of stuff I tested, mostly PD. > >This is due to an error in the startup-code of Manx Aztec C Compiler, which >was present to at least version 3.6. This startup-code opened the >dos.library but did not close it automatically. I always added an appropiate >"CloseLibrary" to my programs. Technically, I don't think this is legal to do from within your program. What happens if you close the library, but the exit code produced by the compiler (located in manx:crt_src/_main.c for the Manx compiler I have) tries to do something with the library? Things that come to mind are the CloseLib() calls that Manx makes to close libraries it has opened. Technically, if you close the library, the program should never use it again because it may have been flushed from memory. To do this correctly, you would have to modify the example _main.c source file provided by Manx and then make that your standard startup call. In real life, just using the dos.library even though your program as closed it earlier is probably no big deal since several other programs are using it at the same time so it is always open, but it is technically not a good idea. >Kristian -dave
dillon@overload.Berkeley.CA.US (Matthew Dillon) (10/20/90)
In article <CEDMAN.90Oct14141233@lynx.ps.uci.edu> cedman@lynx.ps.uci.edu (Carl Edman) writes: >In article <14766@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes: > In a previous article, cedman@lynx.ps.uci.edu (Carl Edman) writes: > >In article <14721@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes: > ......... etc... The bug is in OLDER Lattice & Aztec startup code. Current releases of Lattice and Aztec C both close the dos.library on exit. I must say, though, that the way Aztec handles it is really messy. -Matt -- Matthew Dillon dillon@Overload.Berkeley.CA.US 891 Regal Rd. uunet.uu.net!overload!dillon Berkeley, Ca. 94708 USA -: my *second* computer is built in :-
jep@mtiame.oz (Jesper Peterson) (10/22/90)
In article <CEDMAN.90Oct13225759@lynx.ps.uci.edu> cedman@lynx.ps.uci.edu (Carl Edman) writes: >PS: I always opened and closed not only "dos.library" but even "exec.library". >(Yes, I am a neurotic :-). Considering exec.library "needs" to be open before you can open it I understand how you became neurotic. -- ACSnet: jep@mtiame.mtia.oz UUCP: ...!uunet!munnari!mtiame.oz!jep [...] I had to leave out reality to keep the post clean and to the point. - jeremy@milton.u.washington.edu (Jeremy York) in rec.music.misc
cedman@lynx.ps.uci.edu (Carl Edman) (10/23/90)
In article <302@mtiame.oz> jep@mtiame.oz (Jesper Peterson) writes: In article <CEDMAN.90Oct13225759@lynx.ps.uci.edu> cedman@lynx.ps.uci.edu (Carl Edman) writes: >PS: I always opened and closed not only "dos.library" but even "exec.library". >(Yes, I am a neurotic :-). Considering exec.library "needs" to be open before you can open it I understand how you became neurotic. Really ? You don't mean that the call to OpenLibrary actually is at offset -552d from the base of exec.library and that it actually is in WOM ? Now, please don't offend me or the readers of this group by assuming that anyone does not know this. The practice I described above is a sign of chronic neat-ness , not of complete ignorance. Carl Edman PS: I had hoped that my reference to reading the source to exec.library would have spared me such answers :-). Theorectial Physicist,N.:A physicist whose | Send mail existence is postulated, to make the numbers | to balance but who is never actually observed | cedman@golem.ps.uci.edu in the laboratory. | edmanc@uciph0.ps.uci.edu