file ThreadUnsafe.i3 was added on branch release_branch_cm3_5_8 on 2009-09-12 18:06:10 +0000 (detail)
merge head to release, making them nearly identical for m3core - Tony's thread fixes - add struct_sockaddr_un - just one set of socklen_t wrappers (detail)
just go ahead and make .debs for all targets except NT386 (detail)
put in the retry logic for msi/deb too that tgz has (detail)
.deb creation work in progress; commit it so I don't lose/misplace it again (detail)
delete setup.txt in the loop so it doesn't keep accumulating (detail)
Remove idle thread code. I had recently bumped the idle thread count to 0 and the code is gone in head. However this was in response to heap corruption that was ultimately found to be caused by something else, so maybe maybe maybe we should restore this code for a performance gain. Maybe. (detail)
Add support for Hudson CLEAN variable. Add new test function test_build_system, which first tries a straight forward build of the core system and installs the compiler, and if that fails, tries again using the upgrade.sh script.
If I understood correctly, the last step should not be needed anymore with an up-to-date cm3. (detail)
use ; for portable path list separator when possible -- start fixing the bugs that occur when cygwin Python tries to auotconfigure LIB and INCLUDE, but not done, ok (detail)
remove the absolutely terrible looking nested news frame; leaving the link would be reasonable but is also redundant with the list on the left (really this all needs to be rewritten..) (detail)
replace /dev/null with os.devnull; this should stop us from creating the file nul that causes the makedist task to fail in strange ways, unless I cleanup the workspace between runs (detail)
paths written into .wxs file need to be Win3 paths (again, elego Python is Cygwin..mine is not) (detail)
new files from head, in case we fully merge ThreadWin32.m3 (detail)
remove another unused file (the function is in ThreadF) (detail)
manually merge most changes (in this directory) from head the main point here is to workaround strange bugs in older releases, such as seen in Hudson on PPC_DARWIN where the env variable is missing There was also the "rewrite" of equating SOLgnu and SOLsun. (detail)
rename "cm" lock to "giant" for readability to and to match head (detail)
equate thread id with Windows GetCurrentThreadId this requires moving a few perf notifications to be from within the thread instead of the e.g. creating/joining thread (detail)
.deb creation work in progress; commit it so I don't lose/misplace it again (detail)
path adjustments for msi building, not quite right but might suffice; try to diagnose a strange printout (detail)
heed different workspace layouts on master and slaves (detail)
further accounting for Cygwin python -- double up slashes on command line; and I think print the wierd printout by redirecting stderr to stdout (wierd because it isn't ordered as expected, ok) (detail)
very minor merge from head - drop DEC SRC support (M3BUILD, M3SHIP), unused, probably incomplete, only two lines of code deleted here - very minor formating (roughly another two lines edited) (detail)
don't print an error deleting nonexistant file (detail)
general problem of too much stuff at global scope; workaround some of it (to address that make-msi.py is complaining about findstr against cm3.exe, cygwin path of cm3.exe being passed to win32 findstr) (detail)
(more working around that elego uses cygwin Python but I use Win32 Python) (detail)
a little code to write cm3.cfg based on cm3 -version (detail)
for msi building, hack out executable code from global scope, maybe can structure better later (detail)
fix multiple parts of the date +%s replacement putting [] here is wrong use M3DATE=`pwd`/m3date Maybe we can just remove all of this entirely? (Solaris date doesn't support +%s, so we'd remove the uses, not just the this replacement.) (detail)
