gmane.lisp.scheme.scheme48

no more SIGSEGV when trying to bootstrap Scheme48 from VCS

Ivan Shmakov — 2008-09-30T17:04:01

IS> The development code that crashed was built the way just like the IS> one I've used to build the development code with itself, i. e.: MS> I think I somehow got into a mind state where I believed you're MS> doing all of this on Windows: sorry about that. On what platform MS> are you doing it? IS> I do it on GNU/Linux, as of Debian Etch (amd64 port.) MC> Ivan, do you still see this problem with the current development MC> sources (5f5e619749e0)? I've just built initial.images of f6856f11c6f34 with 6ab92621cc7c, and then again, with itself -- it built just fine, with the images matching bytewise: $ sha1sum build/initial.image-* a27bb81b8ceea2fe999382e8e80d15ae60894e11 build/initial.image-32 a27bb81b8ceea2fe999382e8e80d15ae60894e11 build/initial.image-32.old 658ba49dc83a15fce804c0cab19d623d35dd2eb8 build/initial.image-64 658ba49dc83a15fce804c0cab19d623d35dd2eb8 build/initial.image-64.old $ Thanks! I still wonder, how difficult would it be to make initial.images building a part of

[Scheme Steering Committee announcements] Scheme Language SteeringCommittee election: preliminary announcement

Mitchell Wand — 2008-09-24T05:16:44

_______________________________________________ Scheme-announcements mailing list Scheme-announcements< at >lists.ccs.neu.edu https://lists.ccs.neu.edu/bin/listinfo/scheme-announcements

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Marcus Crestani — 2008-09-22T21:39:02

IS> I do it on GNU/Linux, as of Debian Etch (amd64 port.) Ivan, do you still see this problem with the current development sources (5f5e619749e0)? I was not able to reproduce the error on either Linux galibier 2.6.18-92.1.6.el5 #1 SMP Wed Jun 25 12:38:37 EDT 2008 x86_64 x86_64 x86_64 GNU/Linux and Darwin izoard 9.5.0 Darwin Kernel Version 9.5.0: Wed Sep 3 11:29:43 PDT 2008; root:xnu-1228.7.58~1/RELEASE_I386 i386.

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Ivan Shmakov — 2008-09-14T14:54:04

>> Perhaps the runtime and PreScheme compiler are not quite in sync in >> the latest hg? I could be wrong, but it seems like in general one >> would build the development system with a stable version of the >> compiler (unless one were working on the compiler of said >> development system). > In principle, it should work, but it's not the usual mode of > operation of the developers. I'll check it out. Actually, I'd expect that any revision should be able to build itself in full. Besides, isn't such a build a good test for the system? Therefore, I'd suggest to include (re)building of the essential parts (such as initial images and the C code) as part of the Make `test' target. The resulting files should then match byte-wise those supplied with the release. It may make sense to include $ make test as the final step of $ make enough (unless disabled by the user), and to fail on the $ make install stage if such a test wasn't passed (nor disabled by the user.)

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Ivan Shmakov — 2008-09-13T21:48:32

>>>>> How did you build the running Scheme 48 from the development code >>>>> itself? >>>> The same way as I've tried to build the development version of >>>> Scheme48 with itself later (already posted to this thread), only >>>> using Scheme48 1.8 as the basis. [...] >>> So *how* did you build the development version that crashed? :-) >> I guess, I'm now the one confused with the question. >> The development code that crashed was built the way just like the >> one I've used to build the development code with itself, i. e.: > I think I somehow got into a mind state where I believed you're doing > all of this on Windows: sorry about that. On what platform are you > doing it? I do it on GNU/Linux, as of Debian Etch (amd64 port.) As I was suspicious about possible bugs within the new FFI (well, what's the most likely place for bugs which lead to segmentation faults, anyway?), I've build Scheme48 as of 6ab92621cc7c [1]. It wasn't a big surprise that /that/ version did indeed compile i

Re: a Scheme48 wrapper for Sqlite3

Ivan Shmakov — 2008-09-08T09:45:32

IS> I've made some changes to the code, but I'm currently too busy to IS> finish the work. I'm therefore posting the code in its current IS> state in the hope that it might be useful to someone. I've reworked the code a bit, and the user-visible changes are roughly as follows. * The code is now rewritten to utilize the new Scheme48 FFI (for to-be Scheme 1.9.) * The error-signalling variants are now provided for a number of functions; these variants are: `sqlite3-open', `sqlite3-close', `sqlite3-prepare-2', `sqlite3-prepare', `sqlite3-prepared-finalize!', `sqlite3-prepared-reset!', `sqlite3-prepared-step!'. * The interfaces were extended by `sqlite3-column-name' (with error signalling) and `sqlite3-column-name-1' (without), mostly to allow for `sqlite3-exec' to be reimplemented. * The `sqlite3-exec' is now reimplemented on top of the rest of the interface; both `sqlite3-exec-1' and sqlite3_s48_exec () are gone. * An example makef

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Michael Sperber — 2008-08-26T05:38:16

Ivan Shmakov theory.asu.ru> writes: I think I somehow got into a mind state where I believed you're doing all of this on Windows: sorry about that. On what platform are you doing it?

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Ivan Shmakov — 2008-08-25T16:13:41

>>> How did you build the running Scheme 48 from the development code >>> itself? >> The same way as I've tried to build the development version of >> Scheme48 with itself later (already posted to this thread), only >> using Scheme48 1.8 as the basis. [...] >> (Surely not to mean anything related to either Visual C, or Volkov >> Commander.) > So *how* did you build the development version that crashed? :-) I guess, I'm now the one confused with the question. The development code that crashed was built the way just like the one I've used to build the development code with itself, i. e.: --cut: http://permalink.gmane.org/gmane.lisp.scheme.scheme48/2147 -- $ mkdir +build-1 $ cd +build-1 $ ../configure --prefix="$HOME" ... $ make \ ../build/filenames.make \ i-know-what-i-am-doing \ c/scheme48.h \ ../build/initial.image-{32,64} ... --cut: http://permalink.gmane.org/gmane.lisp.scheme.scheme48/2147 -- (And to be sure, I've just used the commands above to bootstrap t

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Michael Sperber — 2008-08-25T06:13:30

Ivan Shmakov theory.asu.ru> writes: So *how* did you build the development version that crashed? :-)

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Ivan Shmakov — 2008-08-24T16:31:53

>>> Could you elaborate on your build setup? Specifically, how do >>> configure/make fit in the setup? (I use the batch files in the >>> build/ directory, described in WINDOWS.txt to bootstrap.) >> The configure & make steps are similar to those in autogen.sh: > How did you build the running Scheme 48 from the development code > itself? The same way as I've tried to build the development version of Scheme48 with itself later (already posted to this thread), only using Scheme48 1.8 as the basis. > (I was probably a bit quick to assume that I knew what "VCS" means. > What does it mean?) It stands for Version Control System; i. e., Mercurial, when it comes to Scheme48. Alternatively, I could've been speaking about ``the source from the (VCS) repository''. (Surely not to mean anything related to either Visual C, or Volkov Commander.)

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Michael Sperber — 2008-08-21T17:20:52

Ivan Shmakov theory.asu.ru> writes: How did you build the running Scheme 48 from the development code itself? (I was probably a bit quick to assume that I knew what "VCS" means. What does it mean?)

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Ivan Shmakov — 2008-08-21T17:03:42

>> BTW, I'm getting segmentation faults while trying to bootstrap >> recent Scheme48 revisions from the Mercurial repository using some >> likewise recent Scheme48 versions. >> E. g., when trying to bootstrap Scheme48 rev. dc9216ef6864 using >> Scheme48 rev. dc9216ef6864, which was in turn built with Scheme48 >> 1.8: >> $ mkdir +build-1 >> $ cd +build-1 >> $ ../configure --prefix="$HOME" >> ... >> $ make \ >> ../build/filenames.make \ >> i-know-what-i-am-doing \ >> c/scheme48.h \ >> ../build/initial.image-{32,64} [...] > Could you elaborate on your build setup? Specifically, how do > configure/make fit in the setup? (I use the batch files in the > build/ directory, described in WINDOWS.txt to bootstrap.) The configure & make steps are similar to those in autogen.sh: $ nl -ba autogen.sh ... 7./configure && 8rm -rf autom4te.cache && 9rm -f scheme48.image build/initial.image-32 build/initial.image-64 c/scheme48.h && 10make build/file

Re: AC_ARG_ENABLE() isn't used quite properly

Marcus Crestani — 2008-08-21T07:52:19

IS> There's a problem with S48_GLIB and some other Autoconf macros IS> used by Scheme48's configure.in currently. Consider, e. g.: I've pushed a fix for `--enable-glib', `--enable-native-code', `--enable-force-32bit', and `--enable-universal-binary'. Thanks for pointing this out, Ivan!

Re: writing negative flonums

Michael Sperber — 2008-08-21T08:05:31

Ivan Shmakov theory.asu.ru> writes: Fixed. (The problem was only in the development code, not in the release.) Thanks for the report!

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Michael Sperber — 2008-08-21T07:00:41

Ivan Shmakov theory.asu.ru> writes: Could you elaborate on your build setup? Specifically, how do configure/make fit in the setup? (I use the batch files in the build/ directory, described in WINDOWS.txt to bootstrap.)

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Michael Sperber — 2008-08-21T06:54:56

"Scott Bell" gmail.com> writes: In principle, it should work, but it's not the usual mode of operation of the developers. I'll check it out.

Re: SIGSEGV trying to bootstrap Scheme48 from VCS

Scott Bell — 2008-08-20T19:37:16

Hi Ivan, Perhaps the runtime and PreScheme compiler are not quite in sync in the latest hg? I could be wrong, but it seems like in general one would build the development system with a stable version of the compiler (unless one were working on the compiler of said development system). Regards, - Scott On Wednesday, August 20, 2008, at 12:56PM, "Ivan Shmakov" theory.asu.ru> wrote:

AC_ARG_ENABLE() isn't used quite properly

Ivan Shmakov — 2008-08-20T19:06:08

There's a problem with S48_GLIB and some other Autoconf macros used by Scheme48's configure.in currently. Consider, e. g.: $ nl -ba s48_glib.m4 1### s48_glib.m4 --- S48_GLIB macro -*- Autoconf -*- 2# serial 1 3dnl 4dnl 5dnl 6AC_DEFUN([S48_GLIB], [dnl 7AC_ARG_ENABLE([glib], 8[AC_HELP_STRING([--enable-glib], 9 [Use the glib event loop])], 10[dnl Check for glib-2.0 11 PKG_CHECK_MODULES(GLIB, glib-2.0, AC_DEFINE(HAVE_GLIB), AC_MSG_WARN(glib-2.0 not found)) ... I. e., the PKG_CHECK_MODULES(GLIB, ...) (and the rest) will be executed whenever --enable-glib or --enable-glib=ARG is given, no matter what the ARG will be. In particular, the GLIB checks will be made whenever --enable-glib=no is given, or the form treated synonymously to it by Autoconf: --disable-glib. (That way, I was really quite a GLIB user for some time.) The proper way to use AC_ARG_ENABLE() would be, e. g.: AC_DEFUN([S48_GLIB], [dnl AC_ARG_ENABLE([g

SIGSEGV trying to bootstrap Scheme48 from VCS

Ivan Shmakov — 2008-08-20T18:56:03

BTW, I'm getting segmentation faults while trying to bootstrap recent Scheme48 revisions from the Mercurial repository using some likewise recent Scheme48 versions. E. g., when trying to bootstrap Scheme48 rev. dc9216ef6864 using Scheme48 rev. dc9216ef6864, which was in turn built with Scheme48 1.8: $ mkdir +build-1 $ cd +build-1 $ ../configure --prefix="$HOME" ... $ make \ ../build/filenames.make \ i-know-what-i-am-doing \ c/scheme48.h \ ../build/initial.image-{32,64} ... cd ../ps-compiler && \ (echo ',batch'; \ echo ',config ,load ../scheme/platform-interfaces.scm'; \ echo ',config ,load ../scheme/rts-packages-64.scm'; \ echo ',config ,load ../scheme/prescheme/interface.scm'; \ echo ',config ,load ../scheme/prescheme/package-defs.scm'; \ echo ',exec ,load load-ps-compiler.scm'; \ echo ',exec ,load compile-twospace-gc-64.scm'; \ echo ',exit' \ ) | scheme48 -h 8000000 && \ mv ../scheme

writing negative flonums

Ivan Shmakov — 2008-08-20T18:38:15

... Doesn't look like anything good at all. Welcome to Scheme 48 1.9T (made by ivan on Thu Aug 21 01:34:41 NOVST 2008) Copyright (c) 1993-2008 by Richard Kelsey and Jonathan Rees. Please report bugs to scheme-48-bugs< at >s48.org. Get more information at http://www.s48.org/. Type ,? (comma question-mark) for help. 0.1 -1 0.1

ANN: Call for registration: 2008 Workshop on Scheme and Functional Programming

William D Clinger — 2008-08-20T02:44:10

The 2008 Workshop on Scheme and Functional Programming will be held * in Victoria, British Columbia, * on Saturday, 20 September, two days before ICFP. Discounted early registration closes Wednesday, 20 August. The workshop program is online at http://www.ccs.neu.edu/home/will/scheme2008/program.html For more details, see http://www.schemeworkshop.org/2008/ Will Clinger workshop chair