Changelog
mlton (20091107) unstable; urgency=low
* gcc-4.4 miscompiles more than just fpclassify. It is unusable.
* gcc-4.1 ICEs so also not an option.
* => use gcc-4.3, despite an off-by-one bug in mips branch generation
* Using -fPIC to compile changes the assembler enough to get past #552341
mlton (20091104) unstable; urgency=low
* Two more regressions disabled on hurd (thread2 and world5)
* Switch mips back to gcc-4.4
* Found the cause of the segfault: fpclassify is broken on mips
* Disable its on linux/mips and use the MLton IEEE work-around code
mlton (20091101) unstable; urgency=low
* The mips saga continues
* gcc-4.4 miscompiles code
* gcc-4.3 outputs bad branch statements
* => gcc-4.1 might work!
* Don't override gcc version on any architecture but mips[el]
* Change dependencies to require gcc-4.1 only on mips[el]
mlton (20091029) unstable; urgency=low
* Bump the heap-size up a bit to 800m (850m for align8) -> fixes s390.
* Without a porter box it's hard to find the cause of mips(el) build failure
* One likely in the gcc changelog is an ABI change in gcc 4.4
* Apparently the packed attribute behaves differently now
* Since the runtime uses this attribute, rebuild it using gcc 4.4 as well
mlton (20091028) unstable; urgency=low
* Added hurd memory-layout definition for mlnlffigen.
* Use gcc-4.4; it works around #552314 on mips.
* Switch to using cdbs and debhelper.
* Appease lintian (hyphens in manual, licence file, empty directories)
* Set the Homepage field
mlton (20091022) unstable; urgency=low
* Added missing hurd variant to mlton-stubs bootstrap library.
mlton (20091021) unstable; urgency=low
* Revised the heap-size script again to ensure 300m free for the OS.
* Got these buildds black-listed for low ram: fano, nautilus, rem
* Changed s390 alignment to 8 to fix Real64 regressions.
* Filled in regression tests for alpha, hppa, mips/el, powerpc, and s390.
* Added option DEB_BUILD_HEAP to override debian/heap-size defaults.
mlton (20091018) unstable; urgency=low
* Apparently, procps needs to be installed to use 'free'.
* The buildds aren't overloaded, they just suck. Revised heap-size.
mlton (20091017) unstable; urgency=low
* Debian buildds are heavily loaded -> ram-slop 0.7 is a bad idea.
* caused thrashing on nautilus (amd64)
* bootstrap troubles from 20070812 on paer (hppa)
* Solution => fixed-heap based on debian target (rules/heap-size)
* 800MB for 32-bit systems
* hppa, i386, kfreebsd-i386, mips, mipsel, powerpc, s390, sparc
* 1400MB for 64-bit systems
* alpha, amd64, ia64, kfreebsd-amd64
* Added hurd-i386 port
* 700MB fixed-heap (hurd can only map 800MB physical RAM)
* Disable regression suite on hurd; there is a bug in signal handling on
this platform (#551470) which causes these tests to hang:
mutex prodcons signals signals2 suspend timeout
* Otherwise all regressions pass
mlton (20091015) unstable; urgency=low
* Rebuild proper packages using the buildd
* Added mips and mipsel architectures (bootstrap already uploaded)
* Switch to debian native version numbers
* Base the MLton version on the changelog
* Mark the smlnj script as bash-specific (closes: #530142)
mlton (20091013-1) unstable; urgency=low
* New maintainer
* New upstream version taken from svn r7263
* Standards-version 3.8.3.0 (no changes)
* Added alpha ia64 kfreebsd-i386 s390 ports
* manual uploads of (half-broken) bootstrap packages will follow
* Handle nostrip build option (closes: #437581)
-- Ubuntu Archive Auto-Sync <email address hidden> Mon, 30 Nov 2009 23:46:08 +0000
