From: Matt C. <cmc...@in...> - 2000-11-23 01:49:28
|
Hi, Before I start digging into this problem, I'd like to make sure I'm not going to discover what someone else has already documented elsewhere. :-) I got the CLOCC tarball and followed the directions in INSTALL: > 1. Set the environment variable LISPTYPE appropriately No problem. LISPTYPE=cmucl, CLOCC_DUMP=cmucl > 2. Edit the logical hosts definitions in the file "clocc.lisp" in the OK. > 3. Type "make clocc-top" in the top-level directory. No problem. Dumps the image and everything. > 4. You should be able to compile any package in CLOCC now by typing Not exactly. > $ cd src/cllib > $ make system > /local/tmp/clocc/clocc/bin/run-lisp -i /local/tmp/clocc/clocc/clocc-top \ > -x '(funcall (intern "COMPILE-SYSTEM" :mk) "cllib")' > ; Loading #p"/local/tmp/clocc/clocc/clocc-top.sgif". > ; Loading #p"/local/tmp/clocc/clocc/src/cllib/cllib.system". > > In: LET (TBC) > (LET (TBC) > (MAKE:DEFSYSTEM CLLIB > :SOURCE-PATHNAME > (TRANSLATE-LOGICAL-PATHNAME "clocc:src;cllib;") > :SOURCE-EXTENSION ..)) > Warning: Variable TBC defined but never used. > make: *** [system] Error 1 I actually am able to make some progress by: o Not dumping the core image o Working through the pieces step by step from Lisp Is there something newer that I should be following for the procedure here, or should I just keep plowing forward and post a summary of what needs to be done to make CLLIB work under CMUCL? (In case anyone cares, I'm using CMUCL18b under IRIX 6.5.) -- Matt Curtin, Founder Interhack Corporation http://www.interhack.net/ "Building the Internet, Securely." research | development | consulting |
From: Sam S. <sd...@gn...> - 2000-11-23 20:28:38
|
> * In message <148...@an...> > * On the subject of "Compiling cllib under CMUCL" > * Sent on Wed, 22 Nov 2000 20:36:28 -0500 (EST) > * Honorable Matt Curtin <cmc...@in...> writes: > > I got the CLOCC tarball and followed the directions in INSTALL: I use CVS. maybe you would have a better luck with it too. > > $ cd src/cllib > > $ make system > > /local/tmp/clocc/clocc/bin/run-lisp -i /local/tmp/clocc/clocc/clocc-top \ > > -x '(funcall (intern "COMPILE-SYSTEM" :mk) "cllib")' > > ; Loading #p"/local/tmp/clocc/clocc/clocc-top.sgif". > > ; Loading #p"/local/tmp/clocc/clocc/src/cllib/cllib.system". > > > > In: LET (TBC) > > (LET (TBC) > > (MAKE:DEFSYSTEM CLLIB > > :SOURCE-PATHNAME > > (TRANSLATE-LOGICAL-PATHNAME "clocc:src;cllib;") > > :SOURCE-EXTENSION ..)) > > Warning: Variable TBC defined but never used. this is not an error. > > make: *** [system] Error 1 what was the last message from CMUCL? I have no problem compiling everything with CMUCL 18b on solaris 2.5.1 please note that some things in cllib require Gray streams, not dumped with CMUCL by default. you will have to get 2 files gray-streams*.lisp from the CMUCL sources, compile and load them. -- Sam Steingold (http://www.podval.org/~sds) Support Israel's right to defend herself! <http://www.i-charity.com/go/israel> Sufficiently advanced stupidity is indistinguishable from malice. |
From: Matt C. <cmc...@in...> - 2000-11-24 04:04:02
|
>>>>> "sds" == Sam Steingold <sd...@gn...> writes: sds> I use CVS. maybe you would have a better luck with it too. I grabbed the 11/23 snapshot by FTP. (CVS is difficult to do right at the moment...) sds> what was the last message from CMUCL? Everything that I posted is exactly what gets shown. I worked with it interactively to find where it was dying, and it was failing to define the logical pathname "cllib" in the EVAL-WHEN form. By changing the "cllib:" to "clocc:src;cllib;" whever sources were called, everything would build up to the point of xml.lisp. Using the 99.06.14 build of CMUCL for Solaris 2.5.1, the logical pathname problem is fixed, but it dies at the same point in xml.lisp. Fortunately, this error is much more verbose: Compiling DEFUN XMLNS-GET: * added XML namespace: #S(XML-NAMESPACE :URI "http://www.w3.org/XML/1998/" :PRE "xml" :NHT #<EQUAL hash table, 0 entries {7368C55}>) * added XML namespace: #S(XML-NAMESPACE :URI "" :PRE "" :NHT #<EQUAL hash table, 0 entries {73D18ED}>) Byte Compiling Top-Level Form: Byte Compiling Creation Form for #<KERNEL::CLASS-CELL XML-NAMESPACE>: Byte Compiling Top-Level Form: Compilation unit aborted. 1 note Compilation aborted after 0:00:04. make: *** [system] Error 1 sds> I have no problem compiling everything with CMUCL 18b on sds> solaris 2.5.1 I should also point out that this is also on a sun4u, just in case that makes any difference. sds> please note that some things in cllib require Gray streams, not sds> dumped with CMUCL by default. you will have to get 2 files sds> gray-streams*.lisp from the CMUCL sources, compile and load sds> them. Ah ... OK ... I got gray-streams-class and gray-strams from the CMUCL sources, compiled, loaded, and dumped them with clocc-top.core. They loaded find, but the compile of xml.lisp still pukes in the same place. -- Matt Curtin, Founder Interhack Corporation http://www.interhack.net/ "Building the Internet, Securely." research | development | consulting |
From: Sam S. <sd...@gn...> - 2000-11-24 04:53:52
|
> * In message <148...@an...> > * On the subject of "Re: Compiling cllib under CMUCL" > * Sent on Thu, 23 Nov 2000 22:50:57 -0500 (EST) > * Honorable Matt Curtin <cmc...@in...> writes: > > >>>>> "sds" == Sam Steingold <sd...@gn...> writes: > > sds> please note that some things in cllib require Gray streams, not > sds> dumped with CMUCL by default. you will have to get 2 files > sds> gray-streams*.lisp from the CMUCL sources, compile and load > sds> them. > > Ah ... OK ... I got gray-streams-class and gray-strams from the CMUCL > sources, compiled, loaded, and dumped them with clocc-top.core. They > loaded find, but the compile of xml.lisp still pukes in the same > place. Oh--- I am sorry, now I remember this! I hate to say that, but I believe that this is a bug in cmucl. I did report it to a cmucl list (on 2000-07-20), but got no reply. I lack intimate knowledge of the cmucl internals to debug this. sorry. if you can isolate the problem to a small reproducible crash, you should post it to the cmucl list and we all would appreciate it. -- Sam Steingold (http://www.podval.org/~sds) Support Israel's right to defend herself! <http://www.i-charity.com/go/israel> Small languages require big programs, large languages enable small programs. |
From: Matt C. <cmc...@in...> - 2000-11-24 13:35:56
|
>>>>> "sds" == Sam Steingold <sd...@gn...> writes: sds> I hate to say that, but I believe that this is a bug in cmucl. I believe you're right. I just thought it would be nice to get a second opinion. :-) For the time being, I'm using CLISP so we can keep moving, and we're continuing to look at the problem in CMUCL. (The good news is that our code sticks pretty strictly with ANSI wherever possible, and aside from some socket interface stuff, which turned out to be easily replaceable with the stuff from the CLOCC port package. How cool is that!) As it turns out, if we use a very recent build of CMUCL under Linux, it works without any problems. 18c has just been tagged, so we'll try it with an official 18c build and report our success. If it has not been fixed, I'll see if I can come up with a bug workaround that isn't obnoxious. Thanks! -- Matt Curtin, Founder Interhack Corporation http://www.interhack.net/ "Building the Internet, Securely." research | development | consulting |
From: <jm...@st...> - 2001-01-16 01:11:38
|
Matt Curtin <cmc...@in...> writes: > Hi, > > Before I start digging into this problem, I'd like to make sure I'm > not going to discover what someone else has already documented > elsewhere. :-) I got cllib to build under 18c. Something thing that caused me alot of trouble was this: (defvar *clocc-root* #-(or win32 mswindows) "/home/jmadams/packages/lisp/clocc" #+(or win32 mswindows) "c:/gnu/clocc/" "*The root CLOCC directory.") I neglected to have a trailing / in the value for *clocc-root*. I initially had some of the exact same symptoms that you reported. -- John M. Adams |