2003-06-14 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/batch-driver.scm, /cvsroot/chicken/chicken/build.scm, /cvsroot/chicken/chicken/chicken.h, /cvsroot/chicken/chicken/c-platform.scm, /cvsroot/chicken/chicken/csi.scm, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/doc/README, /cvsroot/chicken/chicken/doc/README.CVS, /cvsroot/chicken/chicken/doc/README.txt, /cvsroot/chicken/chicken/extras.scm, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/lolevel.scm, /cvsroot/chicken/chicken/Makefile.am, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/runtime.c:
- extras.scm: the first argument to `read-string' is now optional
- runtime.c: interning of empty string referenced uninitialized memory
- posix.scm: I/O on ports returned by `process' is now fully nonblocking
- posix.scm: `errno/wouldblock'
- files loaded via the `-extend' compiler option are also searched in the
current include-path
- `##sys#call-with-direct-continuation' and `##sys#direct-return' (they
might turn out useful)
- lolevel.scm: `unbound-variable-value'
- posix.scm: `file-read' didn't check the buffer argument early enough
- extras.scm: `write-string'
- csi doesn't use the lolevel unit anymore
- `chicken-setup.c' was not up to date
- added BOOTSTRAP_PATH to Makefile.am, this simplifies the invocation
of make when bootstrapping the system from CVS sources
[Thanks to Peter Keller]
2003-06-06 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/runtime.c, /cvsroot/chicken/chicken/tcp.scm:
- tcp.scm: tcp-listener-port
- fixed bug in runtime.c that resulted in fopen() being called with invalid file-mode
- tcp.scm: EWOULDBLOCK wasn't defined under Win32 with MSVC
- the printer mixed up "input" and "output" when printing port objects
- fixed bug in read-char handler of pipe-ports
- fixed two unbound variable bugs in library.scm and posix.scm
- commit updates ChangeLog at website and send message to mailing list.
In other words: SMTP and FTP extensions are imminent ;-)
2003-06-02 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/compiler.scm, /cvsroot/chicken/chicken/csi.scm, /cvsroot/chicken/chicken/doc/FAQ.html, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/doc/README, /cvsroot/chicken/chicken/doc/README.txt, /cvsroot/chicken/chicken/eval.scm, /cvsroot/chicken/chicken/extras.scm, /cvsroot/chicken/chicken/format.scm, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/lolevel.scm, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/runtime.c, /cvsroot/chicken/chicken/scheduler.scm, /cvsroot/chicken/chicken/support.scm, /cvsroot/chicken/chicken/tcp.scm, /cvsroot/chicken/chicken/batch-driver.scm, /cvsroot/chicken/chicken/build.scm, /cvsroot/chicken/chicken/chicken.h:
- fixed problem with scheduler.scm (C_msleep) under Windows
- overhauled I/O subsystem
- lolevel.scm: invalid-procedure-call-handler
- declaration `safe-globals'
- internal compiler for `##core#app'
- eval/visit uses `##core#app'
- debug-option `D'
- (hopefully) settled all thread-safety issues in `format.scm'
- nonblocking I/O for tcp-ports
- made `##sys#errno' an int, instead of an integer (library.scm)
- fixed bug in tcp.scm: partial writes to tcp-port didn't remove sent part
- `##sys#thread-block-for-i/o!' for output (or mixed?) mode doesn't work properly (deactived,
does simple polling)
2003-05-21 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/chicken-setup.scm, /cvsroot/chicken/chicken/doc/FAQ.html, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/scheduler.scm:
- scheduler.scm: timeout-values in nanoseconds where calculated incorrectly
- chicken-setup.scm: -wrap doesn't create registry, if not existing
- ##sys#special-read-syntax-table
- updated FAQ a little
2003-05-19 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/chicken.h, /cvsroot/chicken/chicken/compiler.scm, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/eval.scm, /cvsroot/chicken/chicken/highlevel-macros.scm, /cvsroot/chicken/chicken/moremacros.scm, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/runtime.c, /cvsroot/chicken/chicken/scheduler.scm:
- `define-foreign-variable' isn't seen by `visit' anymore
- `C_flonum()' advanced allocation pointer wrongly on 64-bit platforms
- `parameterize' accepts arbitrary expressions in the parameter position of the binding-list
- applied patches by Sven Hartrumpf that remove unneeded special casing for ICC
- constants defined via `define-constant' are visible inside constant-expressions for
that form
- the scheduler doesn't waste CPU cycles when waiting for thread- or I/O-timeout, unless
other threads are ready [Thanks to Chris Double]
2003-05-08 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/doc/README, /cvsroot/chicken/chicken/doc/README.txt, /cvsroot/chicken/chicken/support.scm, /cvsroot/chicken/chicken/build.scm, /cvsroot/chicken/chicken/chicken.h, /cvsroot/chicken/chicken/compiler.scm, /cvsroot/chicken/chicken/eval.scm, /cvsroot/chicken/chicken/highlevel-macros.scm, /cvsroot/chicken/chicken/moremacros.scm, /cvsroot/chicken/chicken/parameters.scm, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/runtime.c:
- fixed bug in `process-execute', pointers to potentially gc-moved strings where retained
[thanks to Joerg Wittenberger]
- `define-constant' evaluates 2nd argument at compile time
- fixed bug with mutable constants in `define-constant' forms
- visiting handles `define-foreign-type' and `define-foreign-variable'
- fixed compiler bug in support.scm: `estimate-foreign-result-size' allocated one word too little if flonum result
has to be aligned
- `##core#inline[_allocate]' expanded incorrectly with the hygienic macro system
[Thanks to Panagiotis Vossos]
- fixed 32-bit dependencies: C_FLONUM_TAG was wrong on 64-bit platforms; C_equalp() used `int' where a `C_word'
was needed
- 8-byte alignment was incorrectly handled for 64-bit platforms
2003-05-04 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/library.scm:
- `output-port?' returned 0 on closed port [Thanks to Burton Samograd]
* /cvsroot/chicken/chicken/chicken.h, /cvsroot/chicken/chicken/chicken-setup.scm, /cvsroot/chicken/chicken/compiler.scm, /cvsroot/chicken/chicken/csi.scm, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/doc/README, /cvsroot/chicken/chicken/doc/README.txt, /cvsroot/chicken/chicken/eval.scm, /cvsroot/chicken/chicken/extras.scm, /cvsroot/chicken/chicken/format.scm, /cvsroot/chicken/chicken/highlevel-macros.scm, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/moremacros.scm, /cvsroot/chicken/chicken/parameters.scm, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/runtime.c, /cvsroot/chicken/chicken/srfi-13.scm, /cvsroot/chicken/chicken/support.scm, /cvsroot/chicken/chicken/batch-driver.scm, /cvsroot/chicken/chicken/build.scm:
- `substring' signals (exn bounds)
- replaced several occurences of `substring' with `##sys#substring'
- moved the description of `process' into the proper manual section
- `process' should read non-blocking, now
- buffer grows by doubling it's size in `read-line'
- chicken-setup: defaults options are "-O2 -d0"
- format is thread-safe
- the transcript port wasn't properly recognized as a tty-port
- vector-copy! signals (exn bounds)
- several small optimizations in library.scm
- non-exported globals where still exported, if ref'd before definition
- non-blocking `process' seems to work
- `csi -setup-help' and -uninstall
- `string-null?' is expanded inline
- visit handling constant- and inline-definitions
- `define-constant' wasn't consistently handled in hygienic and non-hygienic mode
- removed procedure-definition syntax for `define-constant'
2003-04-23 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/build.scm, /cvsroot/chicken/chicken/c-backend.scm, /cvsroot/chicken/chicken/chicken.h, /cvsroot/chicken/chicken/chicken-setup.scm, /cvsroot/chicken/chicken/compiler.scm, /cvsroot/chicken/chicken/configure.in, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/doc/README, /cvsroot/chicken/chicken/doc/README.txt, /cvsroot/chicken/chicken/eval.scm, /cvsroot/chicken/chicken/extras.scm, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/lolevel.scm, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/regex.scm, /cvsroot/chicken/chicken/runtime.c, /cvsroot/chicken/chicken/support.scm:
- hidden globals where not properly checked for boundedness
- the foreign return types `c-pointer' and `pointer' did not return `#f' on NULL results
[thanks to Peter Wang]
- `##sys#void' is handled better by `canonicalize-begin-body' (support.scm)
- fixed space leaks in `regex.scm'
- `read-line' accepts optional limit
- `-setup' gives useful error-message on non-UNIX platforms
- there where two conflicting versions of `provided?'
- chicken-setup handles single-file case
- some warning-options are passed to CFLAGS
- fixed typo in posix.scm [Thanks to Joerg Wittenberger]
2003-04-19 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/highlevel-macros.scm, /cvsroot/chicken/chicken/runtime.c:
- fixed a couple of 32-bit dependencies [Thanks to Panagiotis Vossos]
- fixed bug in syntax-case version of condition-case
* /cvsroot/chicken/chicken/psyntax-chicken.pp:
removed psyntax-chicken.pp
* /cvsroot/chicken/chicken/examples/ctclsh.scm, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/doc/README.txt, /cvsroot/chicken/chicken/doc/README, /cvsroot/chicken/chicken/support.scm, /cvsroot/chicken/chicken/tcp.scm, /cvsroot/chicken/chicken/runtime.c, /cvsroot/chicken/chicken/srfi-25.scm, /cvsroot/chicken/chicken/csi.scm, /cvsroot/chicken/chicken/highlevel-macros.scm, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/moremacros.scm, /cvsroot/chicken/chicken/optimizer.scm, /cvsroot/chicken/chicken/psyntax.scm, /cvsroot/chicken/chicken/batch-driver.scm, /cvsroot/chicken/chicken/build.scm, /cvsroot/chicken/chicken/c-backend.scm, /cvsroot/chicken/chicken/chicken.1, /cvsroot/chicken/chicken/chicken.scm, /cvsroot/chicken/chicken/compiler.scm, /cvsroot/chicken/chicken/c-platform.scm, /cvsroot/chicken/chicken/Makefile.am:
- fixed bug in c-backend.scm: heap-limit settings [Thanks to Sven Hartrumpf]
- fixed bug in runtime.c: handling of -:t... was bogus [Thanks to Sven]
- the default output-file for `chicken' is now the original-filename (+ ".c") [Thanks to Panagiotis Vossos]
- syntax-case uses `##sys#void' in expansions
- removed `error-handler'
- diversified exceptions a little
- srfi-25: signals more sensible exn types
- added `condition-case'
2003-04-08 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/compiler.scm, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/eval.scm, /cvsroot/chicken/chicken/examples/prolog.scm, /cvsroot/chicken/chicken/examples/schelog-support.scm, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/match-support.scm, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/regex.scm, /cvsroot/chicken/chicken/runtime.c, /cvsroot/chicken/chicken/scheduler.scm, /cvsroot/chicken/chicken/script-utils.scm, /cvsroot/chicken/chicken/srfi-13.scm, /cvsroot/chicken/chicken/srfi-14.scm, /cvsroot/chicken/chicken/srfi-18.scm, /cvsroot/chicken/chicken/srfi-1.scm, /cvsroot/chicken/chicken/srfi-37.scm, /cvsroot/chicken/chicken/srfi-4.scm, /cvsroot/chicken/chicken/syntax-case.scm, /cvsroot/chicken/chicken/tcp.scm:
- eval.scm: `##sys#environment-is-mutable' isn't hidden anymore.
- scheduler.scm: ##sys#fetch-and-clear.
- simplified some code in tcp.scm.
- `(build-platform)' returns 'intel for ICC.
- removed `interrupts-disabled' declaration.
2003-04-03 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/configure.in, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/runtime.c:
- Changed calls to AC_INIT and AM_INIT_AUTOMAKE in configure.in to new syntax.
- Recreated autotool files using newest versions.
2003-03-31 flw <flw@ultra.callcc.org>
* /cvsroot/chicken/chicken/batch-driver.scm, /cvsroot/chicken/chicken/build.scm, /cvsroot/chicken/chicken/chicken.1, /cvsroot/chicken/chicken/chicken-setup.scm, /cvsroot/chicken/chicken/compiler.scm, /cvsroot/chicken/chicken/c-platform.scm, /cvsroot/chicken/chicken/csc.bat, /cvsroot/chicken/chicken/csc.in, /cvsroot/chicken/chicken/csi.scm, /cvsroot/chicken/chicken/doc/manual.tex, /cvsroot/chicken/chicken/doc/README, /cvsroot/chicken/chicken/doc/README.txt, /cvsroot/chicken/chicken/eval.scm, /cvsroot/chicken/chicken/highlevel-macros.scm, /cvsroot/chicken/chicken/library.scm, /cvsroot/chicken/chicken/moremacros.scm, /cvsroot/chicken/chicken/posix.scm, /cvsroot/chicken/chicken/support.scm:
- Removed syntax checks from internal `##core#...' forms.
- `require-at-runtime' setup-specification.
- `csi -setup -init' did not install core extensions.
- `seconds->string' (posix.scm) couldn't handle large (but valid) non-fixnum integers
[Thanks to Anthony Carrico]
- `visit' and `-visit' options to csi and chicken.
- `(gc #t)' didn't return number of free bytes.
- `pathname-directory-separator' was #\\ under Cygwin (is now #\/).
- Removed fixnum->flonum coercion warning in reader.
2003-03-23 njbeckford <njbeckford@ultra.callcc.org>
* /cvsroot/chicken/chicken/runtime.c:
Got rid of buffer overflow in C_get_environment_variable.
2003-03-22 njbeckford <njbeckford@ultra.callcc.org>
* /cvsroot/chicken/chicken/Makefile.am:
Undid change that made both MORE_LIBS and MORE_STATIC_LIBS be the same thing.
* /cvsroot/chicken/chicken/Makefile.am:
Made both MORE_LIBS and MORE_STATIC_LIBS be the same thing.
* /cvsroot/chicken/chicken/Makefile.am:
-all-static replaced by -static. On Solaris, need libdl.so and
libnsl.so since no libdl.a nor libnsl.a.
* /cvsroot/chicken/chicken/Makefile.am:
"make doc" and "make testsuites" now use CHICKEN=../chicken
* /cvsroot/chicken/chicken/Makefile.am:
Uses MORE_STATIC_LIBS instead of MORE_LIBS for
lib[xxx_]chicken_la_LDFLAGS so that libtool does not add "-ldl"
whenever something links with lib[xxx-]chicken.la.
* /cvsroot/chicken/chicken/chicken-config.in, /cvsroot/chicken/chicken/configure.in, /cvsroot/chicken/chicken/csc.in, /cvsroot/chicken/chicken/Makefile.am, /cvsroot/chicken/chicken/runtime.c:
configure.in: Made new variable SHLIBS which has -ldl or -ldld, which will only be used in non -all-static code.
Makefile.am: Use SHLIBS to differentiate MORE_LIBS from MORE_STATIC_LIBS.
runtime.c: Do not even try to use dl.h/dlfcn.h/dlopen/shlopen when in all-static mode.
chicken-config.in, csc.in: Use MORE_STATIC_LIBS (%morestaticlibs%) in static mode.
aptly called 1_4_2_01, or, as we call it, sun-jdk14-2.1.
Bugs fixed in this release, according to http://java.sun.com/j2se/1.4.2/ReleaseNotes.html
- DPI setting not supported in GTK L&F
- j2re-1_4_2-iftw.exe installer has problems if IE proxy config is incorrect
- enable JavaUpdate in 1.4.2_01
- Remove \r\rn characters from Balloon Tip text
- JavaUpdate shows 'Java Update not completed' message when reboot is accepted
- Can't input character into the applet text field when doing the applet demo
- Random date/time attributes produced in jars through unpack breaks Java Update
- No warning/error messages if patching fails through incremental update
- The auto download URL in Mantis FCS demo is not correct.
- Trigger installer to reboot if Sun VM is running in IE during Java Update
- Java Update tab should be hidden from non-administrative users
- Web bundle installer fails to install if proxy requires authentication
- Installer needs a more explicit message than "creating jar files"
- Updated PhoneHome behavior for consumers
- Need to update java logos in Java2Demo
- Upgrade from 1.4.2 to 1.4.2_01 & uninstall leaves JavaUpdate registry entries
- JavaScript, wrong url for auto-install
- 1.4.2_01 HTMLConverter refers to the wrong default autodl URL
- jre 1.4.2_01, Fr, computer needs to be rebooted dialog is too long
- jre 1.4.2_01, It, string 'Finishing installation...' is too long
- j2sdk 1.4.2_01, ja, extra '\' character in the installshield
- FloatingDecimal throws an assertion error with -enablesystemassertions flag.
so add the appropriate buildlink. Doesn't affect NetBSD, but makes this
work under Irix. Apparently this is a change I forgot to commit a long time
ago.
support; in particular don't use the C compiler within the bootstrap
ADA installation
(gcc installs "fixinclude" headers which are not portable across
OS versions)
fixes PR pkg/22470 by John R. Shannon
- Include bzip2's buildlink2.mk to work-around different API versions.
- Include db's buildlink2.mk on non-NetBSD platforms to get a similar
feature-set on a pkgsrc platforms.
-current kernel from at least 2003-08-09. Previous 1.4.1 versions
never really worked all that well and needed special kernel options
as well as a -current kernel from that date, so updating to 1.4.2
is better than keeping both.
A complete list of changes from previous 1.4 JDKs is available at
http://java.sun.com/j2se/1.4.2/changes.html
sun-jre13-1.0.9 and sun-jdk13-1.0.9). A complete list of changes is
available at http://java.sun.com/j2se/1.3/ReleaseNotes.html
Bugs fixed in this release:
readObject in java.net.URL fails to restore transient fields
java_g -version throws Segmentation Fault in Linux
Solaris/Intel: Synch test hangs trying to get to compiled code safepoint
LINUX: process reaping is broken
BigInteger(String, int) does not fail on multiple minus signs
JCK1.4, api/java_net/... tests hang in "-Xprof -verbose -verify" mode, linux
JVMPI_MONITOR_WAITED value returned is wrong
Java HotSpot(TM) Client VM warning: Attempt to allocate stack guard pages failed
JDK Downgrade Install: Java Plugin 1.4 Fatal Error
incorrect localization for javac diagnostic in J2SE 1.3.1
JDK 1.3.1_07/1.4.1_02 - New ORBs create new threads forever
1.3.1_03] Infinite loop seems to happen in phaseX.cpp with -server
JNI GetArrayElements fails with zero length arrays.
1.3.1 Escalation: JFileChooser: Access key Alt+o doesn't work for "ok" button
1.3.1 Escalation: JFileChooser: Key not working
1.3.1: Fix CDRInputStream.java optimization code to avoid chunk overflow errors
Right borders missing in JMenuItem and JScrollbar in 1.3.1_0x
behaviour of method changes after dynamic compilation occurs
JVM crashes w/ SEGV when trying to execute Java code > 64 kB
JVM crash while reading FVD described class
JVM crashes trying to draw very small Arc2D pie fraction when antialiasing is on
Backward compatibility - jni apps need to dereference bogus returned reference
[1.3.1_04] JVM Crashes
Bottom border missing in JComboBox.
Applies to 1.3.1 only - ServerSocket's accept() throws unexpected SocketException
Tyler R. Retzlaff in PR pkg/22429.
A complete list of changes is available at
http://java.sun.com/j2se/1.4.1/ReleaseNotes.html but appears to consist of
two bug fixes:
- URLConnection openConnection with setUseCaches(true) fails (jpg, gif)
- Java webcams does not work on JRE1.4+
Most user visible change is: this version of gcc is now installed
in its own directory hierachy.
Else and among other things:
- remove build dependency on bison.
- install info files.
- remove per OPSYS PLISTs.
- example mk file style changed to match with other gcc packages.
- ...
Bump PKGREVISION.
it can fix -lreadline to -ledit if needed. But..this is referencing
-L/usr/pkg/lib/... in it's link for -lrs which then gets transformed and
library not found. Fix by copying librs.a into buildlink's lib dir after
setting up install-base. Also pass MAKE_ENV along to the shell make to then
fix readline vs edit.
NEWS FROM 2.1.3 TO 2.1.4
o Fix bugs treating old objects that have already been considered by
the incremental GC (and survived it). Example:
ObjectMemory globalGarbageCollect.
HomedAssociation class instanceCount
gave 0 instead of 1. As a result, --enable-checking now can be
used.
o Fix bugs when doing #become: between old objects, exactly one of
which has not been considered by the incremental GC and was
incorrectly swept when the collector finally reached it.
o Fix bugs when garbage collection triggered finalization while a
primitive was being run. Finalization is now done in a separate
Process.
o Fix bugs treating very large objects.
o Fix infinite loop when the big object threshold was set between
the size of survivor spaces and the size of the eden.
o Printing Integers was unbelievably inefficient. Fixed together
with some more low-hanging fruit.
o SequenceableCollection>>#replaceFrom:to:with:startingAt: allows
again that stop=start-1 (like replaceFrom: 1 to: 0 with: ...)
Tested with -current on i386 only so far; 3 of the selftests still
failing, for reasons NetBSD cannot be blamed for.
CVS----------------------------------------------------------------------
From the release notes:
This release is unfortunately a bit late and incorporates a whole
bunch of very different changes. The biggest part is an update of the
libraries to incorporate the string types from module `Object' into
ADT:*, IO:*, OS:*, URI:*, and (partially) XML:* modules. The internal
encoding of Unicode strings has been switched to UTF-16. Support for
the MacOS X and Win32 platforms has been improved.
There were several changes to the symbol file format. Please remove
any old symbol files that are still around.
__________________________________________________________________________
(Object, ADT:StringBuffer): Internal representation of Unicode strings
is now UTF-16. `String32' is gone and has been replaced with
`String16'. MAX(UCS4CHAR) is 10FFFFX, the largest possible Unicode
code point.
(Object:BigInt): Renamed string base constructors to follow the naming
conventions from module `Object'.
(ADT:ArrayList, ADT:Dictionary): Have been converted to parametric
types. One known side-effect of this is that a type guard
`Dictionary.Dictionary' can no longer be applied to instances of
`Object.Object'. The reason for this is that the run-time type
information does not include information on type parameters and
therefore cannot test for particular type arguments. (The necessary
run-time information may be added later.)
(ADT:Storable, ADT:Object:Storage): The parameter type of the methods
ReadObject() and WriteObject() has been changed back to
`Object.Object'. The methods now handle instances of `Object.String'
as well.
(IO:Address): `UnknownHostException' is now derived from `IO.Error'.
(IO:Buffer): Introduce method `ReadLineLF()', to read a single line of
8-bit characters from a channel, including the termination Ascii.lf
character.
(IO:FileChannel): File name arguments are now of type STRING. They
are converted to sequences of bytes using OS:Path.Encode(). Dito for
system() from OS:ProcessManagement.
(IO:Socket, IO:SocketChannel): Non-blocking connect is implemented.
(OS:Path): Changed functions to work on STRINGs. Changed function
names to follow the Oberon naming conventions. Dito for GetCwd() from
OS:ProcessParameters.
(Exception): Fixed a bug in PopContext(), which lead to stack
corruption if more than one TRY context is popped from the stack.
Also some minor cleanups.
(HashCode): Collects all hash functions for the various basic data
types.
(URI, URI:*): The URI modules use the string types from `Object',
instead of defining their own variant of this theme. Error signalling
has been converted to Exception.ParseError. Please note that the
interface of these modules may be in for further changes later. The
URI changes also propgate into the XML modules.
(oo2c): New command --package-installed (tests if a package is
present) and --list-packages (list all installed packages).
(oo2c): The "Pascal" calling convention, and the CSTRING and ALIGNX
flags have been ported to v2 by Stewart Greenhill. This is primarily
used for native Windows DLLs.
(oo2c): There is now support for MaxOS X style "framework" libraries
(contributed by Tim Teulings).
(oo2c): The documentation strings of a module have been removed from
the symbol file `M.Sym', and are now stored in a separate file `M.Doc'
(in the same directory as `M.Sym'). Compilation commands only read
the `M.Sym' part. Also, compilation of huge interface modules is now
much faster.
(oo2c): Constant folding on floating point expressions is now enabled.
(oo2c): Various minor bugs with parametric types have been fixed.
around.
User visible changes:
- Install in ${GCC_PREFIX} for all platforms.
- Install example mk.conf in ${PREFIX}/share/examples/${PKGNAME_NOREV}/mk.conf.
Rewrite the Makefile to adopt a look and feel similar to those of
lang/gcc{,3}: convert to USE_NEW_TEXINFO, USE_PKG_INSTALL for FILE_SUBST,
mostly dynamic PLIST, platform independent MESSAGE, ...
Bump PKGREVISION.
PKGNAME if we are installing to an alternative location so that the
pkg system doesn't complain about conflicts.
Now it is possible to install a gcc3 tree to some location where
it can be used to bootstrap another instance of itself...
work on mips and powerpc at least.
See http://gcc.gnu.org/bugzilla/show_bug.cgi?id=9091.
The fix therein (which was applied to the 3.3 branch does not really fix
the underlying problem, and is for powerpc only.
While I'm not an ada expert and this pkg wasn't used for more than
bootstrapping itself, I'd consider gnatpsta dispensable.
has allowed these packages to work :)
set CC, CXX, CPP and F77 in gcc{,3}/buildlink2.mk.
you now only need to define USE_GCC2 or USE_GCC3 to use the pkgsrc
gcc of your choice.
the share/examples/gcc/mk.conf and gcc3/mk.conf files are no longer
installed.
bump PKGREVISIONs.
Apply http://gcc.gnu.org/bugzilla/show_bug.cgi?id=10955
* unroll.c (unroll_loop): Fix off-by-one bug.
Fixes build issue with xdm and -O3
Suggested by fredb
-${prefix} is the same like ${GCC_PREFIX}, but the former is already there
are not for the host compiler, but for the stage-1 "xgcc", which is
not subject to buildlink's redirection.
(The gcc libs are needed because ld is redirected.)
-USE_NEW_TEXINFO, saves some patches
-symlink libiberty.a into build tree instead of buildlink/lib,
saves some patches
-put common make rules into Makefile.common (for all) and
buildaddon.mk (for all but the base "c" pkg)
