Commit Graph

297 Commits

Author SHA1 Message Date
sommerfeld ded2509e8b Missing bit of last change 2001-01-01 15:50:23 +00:00
sommerfeld 3274266e5f Delete forceSerial; just use compatMake for that.
Add -N flag to *really* not execute any commands (useful when using
the -d flags to debug usr/src/Makefile)
Document -N
Update documentation of -n to mention that it still executes commands
for targets marked .MAKE so that the -N/-n distinction is clear.
2001-01-01 15:47:37 +00:00
sommerfeld 1ef47ae656 Include token counts in trace output.
Don't let tokensFree go negative.
2000-12-30 16:38:22 +00:00
sommerfeld dd9c50102d Fix token accounting for .BEGIN/.END jobs 2000-12-30 15:58:34 +00:00
sommerfeld 663ff035b1 Avoid token leaks when aborting after an error. 2000-12-30 14:21:22 +00:00
sommerfeld 129b8c6309 strdup tracefile so that it doesn't get smashed. 2000-12-30 02:52:03 +00:00
sommerfeld 9984263e63 Do lazy token release - don't return tokens to the pipe unless we're
waiting for something or we're done.
2000-12-30 02:51:21 +00:00
sommerfeld 8a39635d34 Back out debug cruft. 2000-12-30 02:06:42 +00:00
sommerfeld a804073cac Add a token-passing scheme to allow a recursive make to successfully
use -j; all make's in a recursive build cooperate to limit the total
number of jobs, using a token-passing scheme.

The current token passing algorithm is similar to the one implemented
by gmake; there is a single pipe which is inherited through the entire
process hierarchy; tokens are obtained by reading a byte from the
"read end" of the pipe, and are returned by writing them to the "write
end".  This exact algorithm is likely to change in the future.

Implementation details:
 - Use the new trace facility to allow measurement of the
effectiveness of different token-passing schemes
 - Get a token in MakeStartJobs(), return it in Make_Update()
 - Eliminate Job_Full() and the jobFull global since they are
redundant with token system.
 - Add an "internal" -J option (to pass the token pipe fd's down to
submakes) and a -T option for tracing.
 - Change how compatMake is forced so that -j means something when
inherited by submakes.
 - When waiting for a token, poll the token-passing pipe as well as
the output pipes of existing jobs.
2000-12-30 02:05:20 +00:00
sommerfeld 694b899b9c Quick and dirty trace mechanism to make it easier to look at how different
parallel make job distribution algorithms work in practice.
2000-12-29 23:11:08 +00:00
mycroft b91d3efb9a Make the fds[] realloc O(n). Also make the rethreading a lot simpler. 2000-12-05 21:57:20 +00:00
sommerfeld e65a50bf34 Boolean consistancy (use TRUE, not 1) 2000-12-05 17:07:01 +00:00
sommerfeld e3f68e5130 Make clearfd() take O(1) time instead of O(N) by moving only the last
job in the arrays to the hole made by the now-dead job.  No
measureable performance difference for -j4, but the code is simpler
this way.
2000-12-05 15:28:55 +00:00
sommerfeld b5e2403ec9 correct performance regression of recent change from select() to
poll() for parallel make:
 - Make the poll() code behave more like the select() code: sleep for
a bit waiting for output rather than busy-wait (eww).
 - Install a no-op SIGCHLD handler so that poll/select wake up early
(with -1/EINTR) when a child exits.
 - Change the default sleep time from 500ms to 5 seconds since we now
wake up promptly when a child exits.
2000-12-05 15:20:10 +00:00
christos fd8f9c7b9f oops forgot to commit this one. 2000-12-04 20:13:29 +00:00
christos 8ba1720b96 mycroft did not like my O(n^2) lookup algorithm, so I made things much
more complicated.
2000-12-04 17:45:17 +00:00
christos 105cc20925 1. switch to using poll by default; can be overriden by compiling with
-DUSE_SELECT
2. make sure that we don't overrun our allocated fd_set if USE_SELECT is defined
2000-12-03 02:19:32 +00:00
christos 709d0cf4cb add noreturn attribute to the functions that need it. 2000-12-03 02:18:14 +00:00
christos 9a19da76d7 1. the tfile patch was incorrect. If we are doing multiple jobs, we
try to open tfile many times, passing bad strings to mkstemp
2. remove extra semicolon after the MESSAGE macro
3. more error checking
4. be more careful about setting things to NULL after freeing.
5. fix a comment that does not apply anymore
2000-12-03 01:27:03 +00:00
christos 6a9332cff1 use 6 X's instead of 5 for portability. 2000-12-03 01:18:15 +00:00
is 135600f947 More format string cleanup by sommerfeld. 2000-10-11 14:46:00 +00:00
christos 89d34c0b70 fixed comment U -> u 2000-09-05 21:08:35 +00:00
christos 73e15c3ea3 :u modifier a'la uniq(1) [from der Mouse] 2000-09-05 17:57:52 +00:00
christos a9e7fd9028 don't attempt to free varNoError too. 2000-08-13 22:47:01 +00:00
christos e8b09532bb PR/10714: SUNAGAWA Keiki: Add newer hpux support, but without removing the
old hpux support.
2000-08-03 15:53:24 +00:00
mycroft d53a63dfff Use a lower threshold for rebuilding hash tables. 2000-06-11 07:54:32 +00:00
mycroft 796fb2ec8c Readd optimization last night. Problems earlier were partially due to the
arguments names on one function being swapped (by a previous author).

Do not do any duplicate suppression when a source list is created.  Instead:
* OP_MADE protects against trying to make the source multiple times.
* A new OP_MARK flag is introduced to suppress duplicates while expanding
  the .ALLSRC variable and .USE targets.
This turns the O(n^2) insertion into O(n) in most cases.

This is tested with a `make build' and some special test cases.
2000-06-11 07:39:52 +00:00
mycroft 83660a50d9 Back out last night's optimization for now. 2000-06-10 22:28:33 +00:00
mycroft 04af209e45 Don't use OP_MARK when attaching suffix transformations. 2000-06-10 22:24:21 +00:00
mycroft edfd0106e2 Introduce an OP_MARK bit, and use it to suppress duplicates during .ALLSRC
and .USE expansion.  Also, remove some more Lst_Member() checks that are now
redundant.
2000-06-10 21:44:08 +00:00
mycroft 83f22c12ea Nuke an `optimization' that made source list creation O(n^2). This may cause
more memory to be used with stupid Makefiles, but it saves a fair amount of
time (~13% just for libc) with sane ones.
2000-06-10 13:48:48 +00:00
sjg 70fc6bfcaf no need for v->name-- on JUNK vars anymore 2000-06-10 05:54:29 +00:00
mycroft b0d51d4783 For VAR_JUNK variables, defer freeing of the name string until the bottom of
Var_Parse(), as it may be used (strdup()ed) by some modifiers.
2000-06-10 04:51:00 +00:00
sjg 5d93c6ecfd Don't free v until after the last use. 2000-06-10 04:17:58 +00:00
mycroft 6bf47769cb Do some evil with VAR_KEEP to make it happier. This seems... wrong. 2000-06-06 09:00:49 +00:00
mycroft 39bf056eb6 Allow further patterns after a :D or :U, as in ODE make. (Copied some code
from VarGetPattern.  It should become more flexible instead.)
2000-06-06 08:44:57 +00:00
mycroft 123757193a Roll back my for-expansion changes. Doing this right is just too hard to be
worth it.
2000-06-06 04:56:52 +00:00
sjg 3d93bb57a6 Check_Cwd_av: Ensure that word after a '(', '{' etc is considered a command,
so that we spot the cd/chdir and don't put one in.
2000-06-06 03:00:11 +00:00
sjg 2293fdd17b Make ::= work again with the VAR_FOR context changes. 2000-06-01 09:39:02 +00:00
mycroft ea218bb6e9 Create a `for' context, and substitute iteration variable from it using the
normal Var_Parse() path.  This allows :R, etc. to work on iteration variables.
2000-06-01 04:16:39 +00:00
sjg cd8908a299 Rats! ${FOO:=bar} is a common usage of the SysV = modifier.
To avoid that, we now do ::[+?!]*= but the SysV = modifier can
conflict with any new modifier.  At there are currently no Makefiles
in our tree that use ${FOO::=bar}
2000-06-01 02:29:21 +00:00
sjg 339f702f71 Implemented a set of assignment modifiers. These solve obscure
problems such as using modifiers on .for loop iterators derived from
local variables (eg .TARGET).
Unless the variable already exists in a global context, these assignments are
local to the current context (this is usually what is wanted).
2000-05-30 02:32:21 +00:00
sjg 489a59f8c6 Expand variables that appear in a [No]Match pattern. Allows
[ x${DONE:M${.TARGET}} != x ] || blah
2000-05-14 15:14:41 +00:00
sjg 214b560694 allow .POSIX to define %POSIX as in FreeBSD 2000-05-11 08:22:40 +00:00
sjg edfc5b4270 Expand variables in variable name passed to Var_Set and Var_Append.
This allows ${FOO}.bar=fubar etc to work.
This does not impact any Makefiles in the current tree but allows
some cool magic (which I need elsewhere).
2000-05-11 07:43:42 +00:00
sjg 0051977fda Pass ctxt along to VarLoopExpand so that local variables such as .TARGET
can be expanded.
2000-05-11 03:32:56 +00:00
sjg 82abb388a6 Revert to the original behaviour wrt sysIncPath, ie if sysIncPath
is an empty list, use DEFSYSPATH.
The current behaviour may have been useful when DEFSYSMK was an
absolute path (hasn't been the case since 1996), but right now
make -m /no/such/dir will fail to find sys.mk and die.
2000-05-10 07:49:35 +00:00
drochner df45d3b6cf support OSF/1 2000-05-04 18:29:53 +00:00
drochner 6d10baf050 conditionalize prototypes for REMOTE support
(seems that REMOTE support is incomplete anyway)
2000-05-04 18:29:12 +00:00
drochner e1710a32d6 make cleaner C for more picky compilers 2000-05-04 18:27:53 +00:00