Show that after finding one RankFINAL segment, traceFindGrey then
goes back to RankEXACT segs before tracing the next RankFINAL segment.
Copied from Perforce
Change: 162626
ServerID: perforce.ravenbrook.com
Switch from DIAG_WITH_PRINTF to _WITH_STREAM_WRITEF.
(Still just from TraceStart).
Copied from Perforce
Change: 162395
ServerID: perforce.ravenbrook.com
trace started. This is just to test the DIAG-out mechanism; it
only affects variety.di.
Copied from Perforce
Change: 162389
ServerID: perforce.ravenbrook.com
Hope to fix in time, but for now I need DIAG-out to work, so switch
to printf (yes, really, I know, sorry).
Also, only do DIAG-out in variety.di, to avoid damage to .ci.
So for now, remove DIAG_WRITEF and DIAG from arenavm and trace.
To test, add DIAG_PRINTF in trace.
Copied from Perforce
Change: 162387
ServerID: perforce.ravenbrook.com
(make initial-size to mps_arena_create comically small,
to prompt VMChunkCreates and occasional MPS_RESERVE_BLOCK: 7's, ie.
commit-limit),
so mps_arena_create asks for 2*testArenaSIZE as before.
Copied from Perforce
Change: 162384
ServerID: perforce.ravenbrook.com
mpm.h: DIAG() et al under control of DIAGNOSTICS;
diag.c: under control of DIAGNOSTICS, plus add header and copyright.
Copied from Perforce
Change: 162214
ServerID: perforce.ravenbrook.com
Also: DIAG-out TraceStart reason, and -- if a non-monor collection --
DIAG-out an ArenaDescribe.
Copied from Perforce
Change: 162213
ServerID: perforce.ravenbrook.com
arena.c: turn off call from ArenaDescribe to GlobalsDescribe,
because it outputs too much stuff (perhaps this change should be
reverted later).
Copied from Perforce
Change: 162212
ServerID: perforce.ravenbrook.com
arenavm.c: add DIAG(ArenaDescribe) in VMChunkCreate;
amcss.c: make initial-size to mps_arena_create comically small,
to prompt VMChunkCreates and occasional MPS_RESERVE_BLOCK: 7's
(== commit-limit, which is ok; probably)
Copied from Perforce
Change: 162211
ServerID: perforce.ravenbrook.com
.\buffer.c(155) : warning C4244: '=' : conversion from 'const int ' to 'char ', possible loss of data
Copied from Perforce
Change: 162169
ServerID: perforce.ravenbrook.com
.\buffer.c(155) : warning C4244: '=' : conversion from 'const int ' to 'char ', possible loss of data
Copied from Perforce
Change: 162167
ServerID: perforce.ravenbrook.com
(plus some edits: integrations were not not verbatim)
poolamc.c: Count the loops of amcScanNailed.
If amcScanNailed looped, ss.unfixedSummary is
not accurate, so move all of the ScanStateSummary into
ss.fixedSumamry, so that trace.c's .verify.segsummary
does not erroneously fail.
This fixes job001548.
trace.c: (comment) explain .verify.segsummary
poolamc.c: AMCSegSketch: four character sketch of seg-state
Copied from Perforce
Change: 162150
ServerID: perforce.ravenbrook.com
when amcScanNailed loops. Highlight cases that would (previously)
have failed .verify.segsummary. Count the loops. Show whether it
wasTotal.
AMCSegSketch: correct it to show stalo and neo the right way round.
Copied from Perforce
Change: 162001
ServerID: perforce.ravenbrook.com
not accurate, so move all of the ScanStateSummary into ss.fixedSumamry,
so that <impl/trace/#verify.segsummary> does not erroneously fail.
See also log file a2nNailedLoopReset.txt.
Copied from Perforce
Change: 162000
ServerID: perforce.ravenbrook.com
shows summary is CORRECT. So: discrepancy between refs seen by
AMCSegAllobjSummary() and those seen by AMCScan() of boarded segment.
Copied from Perforce
Change: 161996
ServerID: perforce.ravenbrook.com
find out what the summary really is, and check against SegSummary().
Initially, check this before each AMCScan. This check does *not* fire
when the unfixed assert does; see new a2kAllobjs.txt logfile.
Copied from Perforce
Change: 161995
ServerID: perforce.ravenbrook.com
Save sketches (abzSketchTotal and abzSketchTotalPrev) in amcseg,
which get Described when unfixedSummary is wrong: let's see whether
there's a pattern to the way the previous summary was created.
Copied from Perforce
Change: 161994
ServerID: perforce.ravenbrook.com
seg.c: SegDescribe: say "buffer: NULL" if it is (rather than saying nothing).
buffer.c: BufferDescribe: now interprets buffer->mode for you.
poolamc.c: amcSegCheck: check Nailboard present onlyif SegNailed for some trace;
AMCSegDescribe: correct SEG_SUPERCLASS (so we see GCSeg fields);
say whether seg is Mobile, Boarded (there's a nailboard), or Stuck;
clearer Map symbols, and print a key for them.
Copied from Perforce
Change: 161983
ServerID: perforce.ravenbrook.com
