bernerus@BIRK.CS.CHALMERS.SE (Christer Bernerus) (08/16/90)
Thanks for the latest patch. It finally got me to build ATK on our sparc's. Here's what I've found so far: 0. Someone didn't compress patch 6 on emsworth. I think this shoul be done. We're only getting about 700 bytes/sec ftp here! 1. Building ATK worked out almost OK, it didn't take so long (maybe bacuse I ran it on a SUN4/490). 1.1. Building atk/ness/doc doesn't work. I got the following message: building (dependInstall) (/ufs/e/andrew/atk/ness/doc) make DESTDIR=/usr/pd/andrew install Checking Destination Directories.... touch install.time (ANDREWDIR=/usr/pd/andrew; export ANDREWDIR; \ CLASSPATH=/usr/pd/andrew/dlib/atk; export CLASSPATH; \ WMHOST=; \ /usr/pd/andrew/bin/nessrun present.n nessauth.d) Starting nessrun (Version 1.3, ATK 15.0); please wait... sh: 19030 Memory fault - core dumped *** Error code 139 make: Fatal error: Command failed for target `nessauth.doc' Current working directory /ufs/e/andrew/atk/ness/doc *** Error code 1 make: Fatal error: Command failed for target `dependInstall' Current working directory /ufs/e/andrew/atk/ness/doc *** Error code 1 make: Fatal error: Command failed for target `dependInstall' 1.2 Building ATK without putting the result in "/usr/andrew" fails in overhead/class/testing since the Imakefile specifies MAKEDOFLAGS=-b ../cmd -g which overrides the standard definition which in my case is MAKEDOFLAGS = ${MAKEDODEBUG} -d ${BASEDIR}/lib -b ${BASEDIR}/bin maybe the Imakefile definition should be MAKEDOFLAGS= -d ${BASEDIR}/lib -b ../cmd -g at least this worked for me. 2. Testing the stuff revealed that messages doesn't work at all: piggy> messages Starting messages (Version 7.14, ATK 15.0); please wait... Illegal instruction (core dumped) 3. Help dies if you do the following: piggy> help compchar Starting help (Version 7.2, ATK 15.0); please wait... Now select the words: Composites File Format a bit into the text. Choose "Show help on selected word" and watch the show. It appears as if you select anything bold and then choose Show help on selected word, help dies immediately. 4. Reading the info about compchar makes me confused: How do I specify an argument to a proc in a *init file ? The "ascii" field in the Composites File Format section isn't described, only specified in the syntax. I may guess it's use since I've been involved a bit in the matter of composed characters, but I bet others will be confused. I'd also have wanted to see more examples in the Quick reference section. Btw. I found out about the extistence of compchar by reading the RCS diffs in patch.08r. I found no file telling me what functional changes have been made to ATK with the 2.1 Mbyte sized patch #6. 5. It seems to me (by reading the manual) that there is no way to specify a default translation of ascii to ATK without having to answer at least one question. We had this possibility with Tom's "swedish" proc's. But now this seems to be gone. Is there maybe another (undocumented) argument to ASCIIToATK that enable me to design a menu entry which does what I usually want to do. I.e. translating a piece of swedish ascii text (with {}|'s) to the real stuff and back again without having to answer silly questions (since I'm always knowing what I'm doing :-) . Well, that's all for now. Chris. bernerus@cs.chalmers.se
bernerus@BIRK.CS.CHALMERS.SE (Christer Bernerus) (08/16/90)
Excerpts from info-andrew: 16-Aug-90 Observations on ATK, PL6. Christer Bernerus@cs.chalmers.se (3302+0) > 4. Reading the info about compchar makes me confused: > How do I specify an argument to a proc in a *init file ? > The "ascii" field in the Composites File Format section isn't > described, only specified in the syntax. I may guess it's use > since I've been involved a bit in the matter of composed > characters, but I bet others will be confused. I'd also have > wanted to see more examples in the Quick reference section. > Btw. I found out about the extistence of compchar by reading the > RCS diffs in patch.08r. I found no file telling me what > functional changes have been made to ATK with the 2.1 Mbyte sized patch #6. Blush! The sole example in cpchar.help described exactly what I wanted, I think I'll RTFM better next time I complain. Nevertheless I miss the description of the "ascii" field and the "changes-since-last-patch" description. > 5. It seems to me (by reading the manual) that there is no way to > specify a default translation of ascii to ATK without having to answer > at least one question. We had this possibility with Tom's "swedish" > proc's. But now this seems to be gone. Is there maybe another > (undocumented) argument to ASCIIToATK that enable me to design a menu > entry which does what I usually want to do. I.e. translating a piece of > swedish ascii text (with {}|'s) to the real stuff and back again without > having to answer silly questions (since I'm always knowing what I'm > doing :-) . Poking around a bit more makes me even more confused: ASCIIToATK enables me to translate a selected text region to ISO characters, but doing the reverse (ATKToASCII) operates on the whole document/inset or whatever. If this is a bug, I think it deserves a fix. If it's a feature, I'd strongly oppose this design. The user interface to ATK is already quite confusing. What you do is that you design procs that appears to be their reversals both by their names and their results. Why do these procs have to have different UI's ?? Suggestion: 1. Modify the UI of ASCIIToATK to include the possibility of specifying a parameter in the .*init file. This parameter should be used as a "go right away and do it" flag. (I did this, it was fairly simple) 2. Use the same UI routines for ATKToASCII. 3. Do this right away before people here in Europe start using this. Otherwise I'm quite pleased with the way you designed the cpchar package. Now we'll go doing some troff hacking -:) Chris. bernerus@cs.chalmers.se ---------------------------------------------------------- Christer Bernerus (SM6FBQ) ! E-mail: bernerus@{cs,utc}.chalmers.se Chalmers University of Technology ! Phone: +46 31 721000 Department of Computer Science S-412 96 Gothenburg, SWEDEN
bernerus@BIRK.CS.CHALMERS.SE (Christer Bernerus) (08/16/90)
Hmm, more info: I built my ATK on a SparcServer 4-490 SUNOS 4.1. ---------------------------------------------------------- Christer Bernerus (SM6FBQ) ! E-mail: bernerus@{cs,utc}.chalmers.se Chalmers University of Technology ! Phone: +46 31 721000 Department of Computer Science S-412 96 Gothenburg, SWEDEN
susan+@ANDREW.CMU.EDU (Susan Straub) (08/16/90)
Excerpts from misc: 16-Aug-90 Observations on ATK, PL6. Christer Bernerus@birk.c (3302+0) > 0. Someone didn't compress patch 6 on emsworth. I think this shoul be > done. We're only getting about 700 bytes/sec ftp here! It's now available as patch.06a.Z and patch.06b.Z ... sorry I missed that. Excerpts from misc: 16-Aug-90 Observations on ATK, PL6. Christer Bernerus@birk.c (3302+0) > 1.1. Building atk/ness/doc doesn't work. I got the following message: ..... This was a bug that occurred in patch.006 (alpha), but fixed in the 'official' release, patch.06a and patch.06b. If you didn't get the 'official' patch 6, you should pick up fix.alp in the alpha-patches directory. If you really do have the official patch and the ness/doc bug is still occuring, please let me know so we can look into the problem. Thanks for your help, Susan Straub Andrew Toolkit Group <susan+@andrew.cmu.edu>
tom@ICASE.EDU (Tom Crockett) (08/17/90)
Excerpts from internet.info-andrew: 16-Aug-90 Re: Observations on ATK, PL6. Fred Hansen@andrew.cmu.e (841+0) > Shell core dumps on Sparcs occur frequently due to a bug in Sun's > version of sh. This is supposed to be fixed by Sun at some time. The usual workaround for this is to unsetenv all but your most essential environment variables before starting the make. I have been successful in the past (haven't tried it with patch 6 yet) by eliminating everything except HOME, SHELL, USER, and PATH, and I set PATH to just /usr/ucb and /usr/bin. Tom Crockett ICASE Institute for Computer Applications in Science and Engineering M.S. 132C e-mail: tom@icase.edu NASA Langley Research Center phone: (804) 864-2182 Hampton, VA 23665-5225
janssen@parc.xerox.com (Bill Janssen) (08/18/90)
Excerpts from ext.andrew: 17-Aug-90 Re: Observations on ATK, PL6. Tom Crockett@icase.edu (820+0) > Excerpts from internet.info-andrew: 16-Aug-90 Re: Observations on ATK, > PL6. Fred Hansen@andrew.cmu.e (841+0) >> Shell core dumps on Sparcs occur frequently due to a bug in Sun's >> version of sh. This is supposed to be fixed by Sun at some time. > The usual workaround for this is to unsetenv all but your most essential > environment variables before starting the make. I have been successful > in the past (haven't tried it with patch 6 yet) by eliminating > everything except HOME, SHELL, USER, and PATH, and I set PATH to just > /usr/ucb and /usr/bin. ATK with patch 6 built for me with an unstripped environment on a SparcStation-1 using SunOS 4.0.3c. Bill