GNU Smalltalk - issues

CPU usage is high when there is an exception

Tue, 03 Nov 2009 08:22:15 -0700

When there is an exception in VisualGST gst uses 100% of the CPU. To reproduce it this is simple in a workspace : self halt

multipart/form-data and Swazoo2.2

Thu, 25 Jun 2009 02:01:46 -0700

multipart/form-data posts don't seem to work with Swazoo 2.2. Go to seaside functional test suite, and try either the multipart input test or the upload test. In both cases the server doesn't respond to the request

extend DLD to support loading Mac OS X frameworks

Wed, 30 Sep 2009 07:13:34 -0700

I get

./genbc ./byte.def ./byte.c ./opt.c ./xlat.c > match.h
/bin/sh: line 1: 60172 Segmentation fault      ./genbc ./byte.def ./byte.c ./opt.c ./xlat.c > _match.h
make[2]: *** [match.stamp] Error 139

gdb says

(gdb) run byte.def byte.c opt.c xlat.c
Starting program: /opt/local/var/macports/build/_Users_damien_var_macports_lang_gst-dev/work/gst-dev-1.0/libgst/genbc byte.def byte.c opt.c xlat.c
Reading symbols for shared libraries +. done

Program received signal EXC_BAD_ACCESS, Could not access memory.
Reason: KERN_INVALID_ADDRESS at address: 0x000000000000007e

awful memory usage with lots of big objects

Fri, 21 Mar 2008 05:43:56 -0700

Big objects are allocated directly in oldspace. The pictures here explain it all: http://shootout.alioth.debian.org/gp4sandbox/fulldata.php?test=pidigits&p1=gst-0&p2=squeak-0&p3=yarv-0&p4=luajit-3 GST goes from the fastest of the bunch (even faster than LuaJIT despite being interpreted) to the slowest.

Class variables only bound in methods following them

Thu, 26 Mar 2009 15:42:29 -0700

Might be by design, but it's confusing… I suppose the first B resolves to a non-existent global ?

Object subclass: Fail [
    a [ stdout << 'a:' << B; nl ]
    B := 'hi there'.
    c [ stdout << 'c:' << B; nl ]
]

st> Fail new a; c
a:nil
c:hi there
a Fail
st>

backtrace shown upside-down

Sat, 31 Oct 2009 07:20:57 -0700

__terminate should be at the bottom, not at the top. This left me puzzled for a while, and seems to be done on purpose. :-) Gwen, what was the rationale?

accept it does not work for debugger and sunit browser

Sat, 31 Oct 2009 07:08:51 -0700

This makes quick turnaround much harder.

RBFormatter>>#acceptMethodNode: output category twice

Wed, 28 Oct 2009 07:52:55 -0700

st> (Array lookupSelector: #print) methodFormattedSourceString
'print [
	"Print a represention of the receiver on the Transcript
         (stdout the GUI is not active)"

	<category: ''printing''>
	<category: ''printing''>
	Transcript show: self printString
    ]'

allow different tools to be opened in the same window

Wed, 14 Oct 2009 05:55:55 -0700

This would allow programming to be done all in the same window, with different tools coexisting and less context-switching on the programmer's part. Probably it would require some refactoring to handle toolbar and menu bar switching.

fileouts and formatting

Tue, 29 Sep 2009 02:55:00 -0700

Fileouts could respect formatting, especially tab/spaces use like gst-converter does.

set default window size for browsers

Mon, 21 Sep 2009 13:06:05 -0700

something around 600x400 for browsers and 1024x600 for the launcher (even if it is started maximized).

Show superclasses from another namespace in parentheses

Sat, 19 Sep 2009 08:36:20 -0700

See the old blox browser for what I mean, and to steal the implementation.

Convenience for working with packages

Fri, 24 Apr 2009 06:55:20 -0700

gst could have an option to load a package, like ruby -r somelib does. Also apparently, using packages requires to install them in some global location first (or ~/.st) but it would be nice if packages could be loaded from the current directory or from besides the image file too. I attach that to VM but a 'tools' components would make sense for issue tracking.

Send Mail from within a NetClients socket handler fails

Thu, 08 Jan 2009 23:36:33 -0700

Script crashing when try and send an email from within a server socket handler. System is GST 3.1 on Linux CentOS 5.2 64 bit. To reproduce... Set up a client and server as per... http://sblinn.jottit.com/gnu_smalltalk_netclients-based_echo_server At http://smalltalk.gnu.org/blog/swoolerton/mail-scripting-0 there is a class called SMTPHelper. To see the fault, set the server atResponse method to something like this... respondTo: aRequest [ | session sessionSocket msg smtpHelper | "sending an email from the respondTo method crashes"

Number comparisons not transitive

Mon, 05 Jan 2009 06:03:18 -0700

See http://bugs.squeak.org/view.php?id=3374 a := 16rFFFFFFFFFFFFF81. b := 16rFFFFFFFFFFFFF82. c := a asFloat. {a = b. a = c. b = c.} gives (false true true ) (should all be false). To fix this one would need to define a separate generality for comparisons and arithmetic, and adjust hashes similarly.

--enable-jit make fails

Mon, 20 Oct 2008 14:35:44 -0700

$ ./configure --enable-jit $ make /bin/bash ../libtool --tag=LIBGST --tag=CC --mode=compile gcc -DHAVE_CONFIG_H -I. -I.. -DDEFAULT_EXECUTABLE=\""/usr/local/bin/gst"\" -I../lib-src -I../lib-src -I../libffi/include -I../libffi/include -I../snprintfv -I../snprintfv -I../sigsegv/src -I../sigsegv/src -I../lightning -I../lightning -I.. -I.. -g -O2 -Wall -Wno-strict-aliasing -Wno-switch -fno-gcse -fstrict-aliasing -Wwrite-strings -Wdeclaration-after-statement -Wno-format -Wpointer-arith -Wno-pointer-sign -MT interp.lo -MD -MP -MF .deps/interp.Tpo -c -o interp.lo interp.c

Add support for PreparedStatements to MySQL and PostgreSQL

Thu, 27 Sep 2007 23:47:57 -0700

Postgres and MySQL support PreparedStatements, but their DBD does not know about them. This requires some refactoring. The first step would be to move parts of the FieldConverter (from the MySQL bindings) up to the generic DBI package.

shared pool environment for compile-time evaluation

Tue, 22 Apr 2008 20:44:31 -0700

There are several places you can put expressions that are evaluated during compilation: # class tag arguments (e.g. <import: SomeVar>) # class var initializers (e.g. MyClassVar := SomeVar) # compile-time evals in methods, and in any of the above (e.g. ##(SomeVar) or MyClassVar := ##(SomeVar) In the VM, all of these are currently managed as compilations on UndefinedObject, with the addition to the pool search of the current namespace (which is twiddled by moving UndefinedObject). This is all very well and recursive, where

Baseless Recursion Crash VM

Tue, 11 Mar 2008 16:57:56 -0700

An infinite loop will crash the VM. Simple Example:


sam@osiris:~$ gst
GNU Smalltalk ready

st> Object subclass: Foo [ Foo class >> new [ ^ self new ] ]
st> Foo new
"Global garbage collection... done"
"Global garbage collection... done, heap grown"
"Global garbage collection... done, heap grown"
"Global garbage collection... done, heap grown"
"Global garbage collection... done, heap grown"
"Global garbage collection... done, heap grown"
"Global garbage collection... done, heap grown"
"Global garbage collection... done, heap grown"



POP client crashes when an email has an attachment
Sat, 01 Mar 2008 12:40:50 -0700
Sample code ...

$ cat ./popTest2.st 
#! /usr/bin/gst  -f
PackageLoader fileInPackages: #('NetClients')!
    | client host username password popClient |
    host := 'mailhost'.
    username := 'mailuser'.
    password := 'mailpw'.
    client := NetClients.POP.POPClient connectToHost: host.
    [client
        username: username
        password: password.
        client login.
        Transcript showCr: 'New messages: ', client newMessagesCount printString.
        Transcript showCr: 'bytes ', client newMessagesSize printString.