(DRJ's changes added PoolGenMessage struct, with text-fields for
gen-name and forward-dest name, and condemned Size)
The mps_message_gc_start() Message, with a _why field, remains.
Copied from Perforce
Change: 160989
ServerID: perforce.ravenbrook.com
#define TractArena(tract). The broken definition worked because
it was always invoked with "TractPool(tract)". (Source-only change)
Copied from Perforce
Change: 160013
ServerID: perforce.ravenbrook.com
MPS: fix Windows build of text-executables needing formats, with new
make-variable FMTTEST, replacing DWTEST and HETEST, and fixing
awlut.exe build. (integrate from version/1.106/...@157962)
(159467)
MPS: fix w3i3mv "all" build with VC6:
1. amsss.c: initialize ambigRoot (uninitialized warning);
2. commpost.nmk:
2a. drop long-vanished protcv.exe;
2b. mpmss.exe needs MVFF;
2c. locusss.exe had no build rule;
3. finaltest.c: main does not use argc,v so declare as main(void);
4. lockutw3.c: add parens to avoid operator-precedence warning;
5. protocol.h: s/interface/interfaceIn: suspected compiler bug.
(integrate from version/1.106/...@158054)
(159468)
MPS: attempt to fix w3i3mv "all" build with VC7:
(unknown if these changes are effective, but they should be safe)
1. mpm.c: +1 to array size to make room for terminator of string initializer;
2. testlib.c: disable warning 4702: unreachable code, for va_end.
(integrate from version/1.106/...@158055)
(159469)
Copied from Perforce
Change: 159471
ServerID: perforce.ravenbrook.com
(unknown if these changes are effective, but they should be safe)
1. mpm.c: +1 to array size to make room for terminator of string initializer;
2. testlib.c: disable warning 4702: unreachable code, for va_end.
(integrate from version/1.106/...@158055)
Copied from Perforce
Change: 159469
ServerID: perforce.ravenbrook.com
1. amsss.c: initialize ambigRoot (uninitialized warning);
2. commpost.nmk:
2a. drop long-vanished protcv.exe;
2b. mpmss.exe needs MVFF;
2c. locusss.exe had no build rule;
3. finaltest.c: main does not use argc,v so declare as main(void);
4. lockutw3.c: add parens to avoid operator-precedence warning;
5. protocol.h: s/interface/interfaceIn: suspected compiler bug.
(integrate from version/1.106/...@158054)
Copied from Perforce
Change: 159468
ServerID: perforce.ravenbrook.com
make-variable FMTTEST, replacing DWTEST and HETEST, and fixing
awlut.exe build. (integrate from version/1.106/...@157962)
Copied from Perforce
Change: 159467
ServerID: perforce.ravenbrook.com
AVER(i < j) on line 465 when you call mps_arena_formatted_objects_walk
when you have objects in an LO pool. (integrate from branch).
Copied from Perforce
Change: 159441
ServerID: perforce.ravenbrook.com
AVER(i < j) on line 465 when you call mps_arena_formatted_objects_walk
when you have objects in an LO pool.
Copied from Perforce
Change: 159440
ServerID: perforce.ravenbrook.com
MPS Wiki: AP User Guide: reword mistake-example to avoid an ambiguity
['may' can mean 'is permitted to', but can also be read as
'might', which can change the meaning].
MPS Wiki: Pool classes: Link to RIT's 1998 guide.mps.pool-classes
Index: Section: About the MPS Wiki
Copied from Perforce
Change: 159391
ServerID: perforce.ravenbrook.com
['may' can mean 'is permitted to', but can also be read as
'might', which can change the meaning].
Copied from Perforce
Change: 159390
ServerID: perforce.ravenbrook.com
modes of use of MPS
gc -- beginner's guide
apguide -- allocation points
glossary
issues with unmananged workspace
Copied from Perforce
Change: 159371
ServerID: perforce.ravenbrook.com
Tidy gc: Complete AP User Guide:
Make AP User Guide consistently .assume.ambig-workspace,
moving notes to APInternals; add Intro and Mistakes;
check example code.
Tidy modes of use: move notes to new unmanaged workspace article.
Copied from Perforce
Change: 159365
ServerID: perforce.ravenbrook.com
Modes of use; Types of object; Types of reference.
And (temporary) still trying to get guarded values right.
Copied from Perforce
Change: 159276
ServerID: perforce.ravenbrook.com
further notes into User Guide and Internals, including initializing
as-yet unused parts of new objects to reduce spoofing.
Copied from Perforce
Change: 159248
ServerID: perforce.ravenbrook.com
Roots -- User's Guide: bare bones: root variants.
Allocation point user's guide: standard reserve-commit
idiom: corrections, and now assume ambiguous reg+stack:
transitioning to this assumption.
Add .talk.RB.2006-06-13 about standard r-c idiom
Add .talk.RB.2006-06-14 about Modes of use of MPS,
and Promises to not flip (single-threaded).
Add .think.RHSK.2006-06-14 about how we could support
unscanned reg+stack, even with multi-threaded, moving
collector.
Copied from Perforce
Change: 159241
ServerID: perforce.ravenbrook.com
Scenario; A mutator with ambiguous references only (to the new
object); How bad is a bogus exact reference?; quick notes on: with
exact refs; flipping at other times.
Copied from Perforce
Change: 159200
ServerID: perforce.ravenbrook.com
