gmane.lisp.scheme.bigloo

-library equivalent of "-Xlinker --whole-archive"

Todd Dukes — 2008-09-04T22:07:35

I have a library that among other things defines a class called graph. My first program creates a data structure that includes instances of this class and writes them to a binary data file. Another program loads this data structure and prints a summary report. This program does not actually need the data in the graph class so there is no code for creating or accessing graph class instances in the code of the program. I use the (library utility) statement in the module clause of my second program to say that it is using the library. At runtime I get an error saying the the class 'graph' cannot be found. I have eliminated the error by putting. (make-graph #f #f) into my summary report program. There should be a better way. I have resolved this in the past with C++ programs by using the --whole-archive flag. Is there a way to do this with libraries in bigloo? Let me know if this makes no sense to anyone and I'll try to clarify. Thanks, Todd.

Re: xml-parse on 3.1b vs. 3.1a

2008-09-04T11:29:42

I am not denying it may be an improvement. I was trying to judge how much code I would have to change to test the beta, and then change back or make versionable when I switched back to the production compiler after testing. Only to change again in a few weeks when I change back to the then new production compiler. I am parsing an xml data file, so for me there is no html. I may try changing the code where I am skipping the header and see if that is the only change affecting me. Somehow my code winds up saying the header is the whole data, but I haven't looked at it. I wanted to know the extent the changes would affect me. It sounds to me from your comments that they may have minor affect on the rest of my code. Thanks, Todd. Cyprien> Hello Todd, I did theses changes because I didn't Cyprien> understand why the first xml declaration wasn't parsed Cyprien> and splitted. I needed it in order to retrieve the xml Cyprien> encoding value, without parse the string given by the Cyprien> i

Re: xml-parse on 3.1b vs. 3.1a

Cyprien Nicolas — 2008-09-04T10:07:54

Hello Todd, I did theses changes because I didn't understand why the first xml declaration wasn't parsed and splitted. I needed it in order to retrieve the xml encoding value, without parse the string given by the instruction symbol. I wanted the xml tag to be parsed like any other html tag, only to have an easy access to the attributes. Sorry but there aren't documented, but I only made changes on the xml tags parsing, not for html. This behavior may be improved, I don't know how is parsed... I suggest to replace the xml-decl by xml, to keep the whole parser coherent with both xml and html tags, as `instruction' isn't a very clear name for all xml declarations. Regards Cyprien 2008/9/3 Todd Dukes freescale.com>:

xml-parse on 3.1b vs. 3.1a

Todd Dukes — 2008-09-03T16:49:14

I see a difference in the way the header information is arranged in the return value of xml-parse between version 3.1a and the 3.1b-beta29Aug08 version. I have copied a small portion of the xml file and the results of printing the parsed data structure below. Are these the only changes I should expect in the parsed xml data structure? If there are any others, are they documented? Thanks, Todd. xml --- xml --- bigloo 3.1b ---- (raw-data ((xml-decl (encoding . UTF-8) (version . 1.0)) (data () ( bigloo 3.1b ---- bigloo 3.1a ---- (raw-data ((instruction . ) (data () ( bigloo 3.1a ----

Re: newbie: wrong (jvm) bytecode generated for constructors

Jakob Praher — 2008-08-26T15:49:02

maybe someone could explain to me, how I know exactly which code generates the bytecodes/expressions for the constructor declarations (methods that start with an underscore in the AST). What about that "-t[2|3|4]" feature - the usage says: "* : only available in developing mode" What is developing mode? When I try to run "bigloo -t -jvm test.scm" I get the following error: *** Warning:bigloo:parse-args No trace for this pass -- ld thank you

newbie: wrong (jvm) bytecode generated for constructors with array signature?

Jakob Praher — 2008-08-25T15:39:52

hi, I give up. After looking at the comptime code for some time, I can't find the missing peaces, to look at the issue all by myself. First of all: I am trying to improve my knowledge of scheme and thus I want to use scheme for gluing together some java components..., by using bigloo -jvm to compile scheme 2 java bytecodes, bigloo version 3.1a on a GNU/Linux machine. So far so good, but: I use the following declarations to boostrap access to java: (module foo (java (class %jobject "java.lang.Object") (class %jstring (constructor new (::string) "java.lang.String") (class %url (constructor new (::%jstring)) "java.net.URL") (array url* ::%url) (class %urlcl (constructor create (::url*)) "java.net.URLClassLoader") ) ... ) ... while the url constructor is good, the urlcl constructor generates wrong bytecode. The outcome of the isntance-of test(generated by rtl_isa) is left on the stack (a constant integer) and that would be used as a parameter to the

Bigloo, snow and make-struct...

2008-08-22T20:15:56

Hi everyone. While porting some code wrote to the Scheme Now (aka snow) system, I stumbled upon what I believe to be a bug in bigloo... The following code shows the behaviour: % cat bigloo-snow.scm ":";exec snow -- "$0" "$< at >" (package* bigloo-snow/v1.0.0 (provide: ;; creates an empty structure type (define-macro* (define-a-struct name) `(define-struct ,name)) ;; appends two symbols (define-macro* (kons a b) (define (symbol-append s1 s2) (string->symbol (string-append (symbol->string s1) (symbol->string s2)))) (symbol-append a b)) ;; calls the creation of a structured element (define-macro* (make-my-struct name) `((kons make- ,name))) ;; same as above ;;(define-macro* (make-struct name) ;;`((kons make- ,name))) )) % cat bigloo-test.scm ":";exec snow -- "$0" "$< at >" (package* bigloo-test/v1.0.0 (require: bigl

Re: bigloo on UltraSparc T1 (Niagara)

Marijn Schouten (hkBst — 2008-08-21T17:34:25

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Manuel.Serrano< at >sophia.inria.fr wrote: I tried the first: unpack gc-7.1.tar.gz and replace its Makefile.direct with the file you attached and install as normal. I get: Emacs configuration: emacs................................. emacs emacs brand........................... emacs22 lisp directory........................ /usr/share/emacs/22.2/site-lisp/bigloo if [ "yes" = "yes" ]; then \ make -C gc boot; \ fi make[1]: Entering directory `/var/tmp/portage/dev-scheme/bigloo-3.1b_beta/work/bigloo3.1b/gc' make -C gc-7.1 /var/tmp/portage/dev-scheme/bigloo-3.1b_beta/work/bigloo3.1b/lib/3.1b/libbigloogc-3.1b.a CFLAGS="-O2 -O2 -mcpu=ultrasparc -pipe -DNO_DEBUGGING -Iinclude - -Ilibatomic_ops-install/include -DFINALIZE_ON_DEMAND -fPIC -I /var/tmp/portage/dev-scheme/bigloo-3.1b_beta/work/bigloo3.1b/lib/3.1b" make[2]: Entering directory `/var/tmp/portage/dev-scheme/bigloo-3.1b_beta/work/bigloo3.1b/gc/gc-7.1' make[2]: *** No rule to make target `/v

bigloo3.1b goes beta

2008-08-21T13:06:41

Hello There, I have just promoted Bigloo3.1b. It has now stepped into its beta stage. My plan is to release it as the next "stable" within a week or so (in any case, no later than the end of the month). In consequence, I would appreciate your help, in particular, in testing the installation procedure. The version 3.1b brings some new features and it also should improve the performance of many compiled programs.

Re: Bee with bigloo3.1b-alpha18Aug08 on Mac OSX

2008-08-21T11:21:56

Hello Rembrandt, I got it. The runtime/Makefile was containing a typo that prevented Bee to be successfully compiled. I will upload a new version in a couple of minutes. Thanks for your report.

Re: bigloo on UltraSparc T1 (Niagara)

2008-08-18T11:32:26

I see on the log you sent that only mach_dep.o is built for SPARC on Linux. I then think that the problem is easy to fix. Could you try something for me? 1- Extract the file gc-7.1.tar.gz that is included in the bigloo/gc directory. 2- Untar it is a temporary directory. 3- Replace the original Makefile.direct with the one attached in this mail and then... 4- follow the installation instructions, Alternative 1 (that is copy Makefile.direct to Makefile and run "make test" *WITHOUT* running configure). Could you tell me if this succeeds? Thanks in advance.

Re: bigloo on UltraSparc T1 (Niagara)

2008-08-18T11:29:33

Yes. Bigloo uses an almost genuine up to date Boehm's GC.

Re: bigloo on UltraSparc T1 (Niagara)

2008-08-18T11:27:38

Hello Marijn, Sorry for this late answer. I have been out of my office for two weeks. Isn't it only a problem of library path? Where is located libgc.so on your Sparc. Is this directory listed in /etc/ld.so.conf?

Re: bignum bit operations?

2008-08-18T11:03:23

Hello Todd, They are not available. May be Jose is willing to add them for the next version...

Re: Bigloo on FreeBSD

2008-08-18T10:54:26

Hello Karsten, Thank you very much for you mail. The latest version no longer includes bglmixer.c. There is actually no need to implement a direct mixer binding for using Gstreamer. This part has thus been removed from the gstreamer binding. Sure. You are right. Changed. I have applied the two patches. I will upload a new version by the end of the afternoon. Thanks a lot for your help.

Re: Bug report

Benedikt Rosenau — 2008-08-12T12:14:36

Re, Ouch, please ignore my last message. I changed the program with the different Schemes and I made a mistake in the parentheses. Benedikt

Bug report

Benedikt Rosenau — 2008-08-12T12:06:35

Hallo Manuel, I attached a short program to this mail. This program leads to a segmentation fault both in interpreted and compiled mode. Compiling with debug information and running through a debugger shows that the problem comes from line 9 and 10 of the source code. However, a backtrace shows that the program ran into an endless loop without reason. Regards, Benedikt (module eliminate-showcase) (define (replace-expression-with-symbol expr sym l) (cond ((equal? expr l) sym) ((list? l) (map (lambda (e) replace-expression-with-symbol expr sym e) l)) (else l))) (define (first-true-sublist l) (let loop ((work-list l)) (if (null? work-list) l (let ((e (car work-list))) (if (list? e) (first-true-sublist e) (loop (cdr work-list))))))) (define (eliminate l) (if (list? l) (let ((sym (gensym)) (sub-list (first-true-sublist l))) `(let ((,sym ,sub-list)) ,(eliminate (replace-expression-with-symbol sub-list sym l)))) l)) (define (fibo n) (if (< n

bignum bit operations?

Todd Dukes — 2008-08-01T20:21:51

Are the bit operations available for big numbers? I am looking for bit-rshbignum and bit-andbignum and do not find them. Am I missing something or are they not available? Thanks, Todd.

Re: bigloo on UltraSparc T1 (Niagara)

Marijn Schouten (hkBst — 2008-07-29T16:27:53

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Manuel.Serrano< at >sophia.inria.fr wrote: I forgot to ask, but the GC that comes with bigloo is just the same as the boehm gc? For some reason I thought it used its own unrelated gc. Marijn - -- Marijn Schouten (hkBst), Gentoo Lisp project, Gentoo ML , #gentoo-{lisp,ml} on FreeNode -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.9 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iEYEARECAAYFAkiPRQkACgkQp/VmCx0OL2zTRQCePtnTAccAW4GIudB08SZuVwNG lKsAoJF5jBOO6jg4ten0fv3agiVXIhTh =YSjE -----END PGP SIGNATURE-----

Re: bigloo on UltraSparc T1 (Niagara)

Marijn Schouten (hkBst — 2008-07-29T16:23:43

Re: bigloo on UltraSparc T1 (Niagara)

Marijn Schouten (hkBst — 2008-07-29T16:00:54

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Manuel.Serrano< at >sophia.inria.fr wrote: Sorry, I should have been more clear. This machine does NOT run Solaris, but indeed runs Linux. boehm-gc compiles on this architecture, but when I try to build bigloo with - --customgc=no on it I get: ** Configure ** Cannot compile gc checker. Something broken with your GC settings. Bigloo cannot be configured with the --customgc=no option. *** ERROR:configure:Standard GC. Aborting You are trying to configure Bigloo so that it uses an existing GC library. The configuration script is enable to make this library working properly. Try to configure Bigloo with "--customgc=yes". The Boehm gc seems to get installed where bigloo expects it by default. However the autoconf/gcstd test fails. When I try to do that test manually: $ gcc testgc.c -o testgc -I /usr/include/gc -lgc /usr/lib/gcc/sparc-unknown-linux-gnu/4.1.2/../../../libgc.so: undefined reference to `GC_save_regs_in_stack' /usr/lib/gcc/sparc-unknown-linux-gnu/4.1.