zmls04@UUNET.UU.NET (Martin L. Smith) (03/20/89)
Originally tried to post this to gnu.ghostscript.bug but that's not locally recognized as a newsgroup. From: zmls04@.uucp () Path: .uucp!zmls04 Newsgroups: gnu.ghostscript.bug Subject: Ghostscript 1.2 on a PC Expires: References: Sender: Reply-To: zmls04@.uucp () Followup-To: Distribution: usa Organization: Amoco Production Company, Tulsa Research Center Keywords: Spent some time trying to bring 1.2 up on an AT with Turbo C V1.5, 1. there was an outright bug in the file containing the arithmetic primivites. The one for `div` would always fail if the second argument (top of stack) was an integer because the test for zero divisor had an (obvious) bug. This bug caused ghost.ps to fail on startup. 2. the compilation produced many complaints (warnings; there were no errors) about pointers to structures of undefined type. These were triggered by references of the form struct HotStuff *HS; where HotStuff was not declared. I presume that this happened on Unix systems as well. I suggest you consider clearing this up. The deluge of messages makes it hard to spot any real problems. 3. gt.exe appeared to work ok. it left the ega's colors scrambled. 4. i ran gs.exe on a short script (copied from the tutorial) that makes three rows of circles across the page. it made the first row and appeared to hang. You might consider: - leaving an .exe version available somewhere (with its support files). - including a test.ps which holds the scripts you used in testing. Sorry this is all complaints. I'm really impressed by the power of this beast and anxious to see it hum. Thanks, from the world.
ghost@SUN.COM (L. Peter Deutsch) (03/22/89)
> Date: Mon, 20 Mar 89 08:51:17 CST > From: parcplace!sun!uunet.uu.net!apctrc!gpe3!zmls04 (Martin L. Smith) > Message-Id: <8903201451.AA28914@gpe3.trc.amoco.com> > To: bug-ghostscript%wheaties.ai.mit.edu > Subject: ghostscript 1.2 > Organization: Amoco Production Company, Tulsa Research Center > > 1. there was an outright bug in the file containing the arithmetic > primivites. The one for `div` would always fail if the second > argument (top of stack) was an integer because the test for zero > divisor had an (obvious) bug. This bug caused ghost.ps to fail > on startup. This bug has already been reported. It'll be fixed in 1.3. > 2. the compilation produced many complaints (warnings; there were no > errors) about pointers to structures of undefined type. These were > triggered by references of the form > struct HotStuff *HS; > where HotStuff was not declared. I presume that this happened on > Unix systems as well. I suggest you consider clearing this up. The > deluge of messages makes it hard to spot any real problems. These messages are a result of the way I get around C's lack of data abstraction. I have an alternative way to handle this that will break lint and some debuggers, but will get rid of the messages. I have a different alternative way to handle it that will slow everything down and create a lot of new compilation dependencies, and I'd rather not use it. > 3. gt.exe appeared to work ok. it left the ega's colors scrambled. Also fixed in 1.3. > 4. i ran gs.exe on a short script (copied from the tutorial) that makes > three rows of circles across the page. it made the first row and > appeared to hang. I typed in exactly the same code (from pp. 69-70 of the blue book) and it worked just fine. Again, though, this is 1.3, with some fixes that haven't made it out to the net yet. > You might consider: > > - leaving an .exe version available somewhere (with its support files). Starting with 1.3, I'll be including gs.exe and gs.map in the fileset, compiled with the most conservative flags (8088 instructions only, and full source line number information). > - including a test.ps which holds the scripts you used in testing. Unfortunately, almost all of my test files are copyrighted material from Adobe publications, which I can legally type in and use myself, but not distribute. I'd be thrilled if someone out there would put together an independent test suite. > Thanks, from the world. Thank *you* for reporting problems and helping make Ghostscript better. - P. D. aladdin!ghost@parcplace.com ...{sun!}parcplace!aladdin!ghost P. O. box 60264, Palo Alto, CA 94306