executed the make job with sigprocmask(SIG_BLOCK) in effect for a
number of signals including SIGCHLD. This caused recursive submakes
of the make process in question not to receive SIGCHLDs when their
jobs exited. This was a second, independent cause of bin/18895 in
addition to the race condition already fixed. Fixed by unblocking all
signals before executing the job.
search list that will be used. Thus 'dot' and 'cur' will appear in
${.PATH} either at the start or end depending on .DOTLAST even though
they are not strictly in dirSearchPath.
When .CURDIR is assigned to - re-set the 'cur' Path.
Finally, when checking subdirs, look in 'dot' and 'cur' (first or last
depending on .DOTLAST) just as we do in other situations.
seconds if a SIGCHLD arrived while make was not blocked in poll(),
by making the SIGCHLD handler write to a pipe included in the poll.
Avoided the need to implement a duplicate fix for the USE_SELECT case
by emulating poll() in terms of select() when USE_SELECT is defined.
Fixes bin/18895.
- consistently support __hpux (which the HP compilers define) as well
as __hpux__ (not sure which compilers set this, but retained anyway)
- fix a typo in the definition of signal(). arguably the codebase should
just be converted to sigaction()...
In these revisions `::' dependency handling was simplified by not linking
the cohort nodes into the dependency graph. This broke dependency checking
on all but the first instance of a `::' target since all of the cohort nodes
now just form a collection of disconnected dependency graphs.
Fix this by keeping a back-reference in each cohort to its leader (the
first instance of a :: node with the same name) and a count of the number
of cohorts that need to be made before dependent nodes are scheduled.
Classically, we'd need six centurions for cohort, but in this case one
suffices...
while setting up a child process. This prevents the major sources of
list inconsistencies that may result from doing list manipulations
from a signal handler.
XXX - all Lst_*() operations are at risk; look for ways to avoid touching
so much global state from the signal handlers.
- don't block the signal we're about to send to ourselves to
take the default action for
- restore the signal handler for the current signal
- remove duplicate sigprocmask() call
As a result, restoring the SIGTSTP handler in JobContinueSig() is no longer
necessary.
1. Compensate for h-pox assert brain damage where it gets confused by the
string in: assert(!memcmp(foo, "in", 2)); I miss the h-pox broken compiler,
I had not had to work around it for years.
2. Oh, finally h-pox has random() and utimes(). We don't need our own anymore.
the removal of .USE nodes from a node's children list in case a given .USE
nodes appears multiple times on that list, preventing the target from ever
making it on the `to be made' queue.
Since the suffix rule processing code deals itself with removing the
transformation nodes from the parents to which it applies them, arrange
for doing the same for .USE nodes in MakeHandleUse() instead of in
Make_HandleUse(), and still use the OP_MARK stuff to avoid duplication
of commands.
Also, since Make_HandleUse() is not a list callback function and its
return value is no longer used, make its return type void.
normal target (the latter dependency being the reason why it's made),
the .MADE parent should not be put on the `toBeMade' list again (in
Make_Update()) since it was already put on it in Make_ExpandUse().
Doing so would cause the rules for the .MADE target to be executed
(at least) twice, and also mess up the unmade children count of _its_
parent, resulting in spurious graph cylce detection.
To achieve this, make sure the unmade children count of the .MADE target
is set to zero before enqueueing it on the `toBeMade' list in
Make_ExpandUse(). Then, in Make_Update(), check if the parent has the
.MADE attribute before diddling with the queue.
For the same reason the status of a node must not be set to UPTODATE
unconditionally in compat mode, since that will prevent the node from
being built even if it is the source of a normal target. Instead,
check both its state and the type of the parent to decide whether to
continue on behalf of the current parent.
* There is no -indent option to .Bd or .Bl, although you would
never know that from its frequent use in this tree. There is a
"-offset indent" combination that makes sense, and you can certainly
say "-width indent".
* Also, you can't markup the -width option argument, tho you CAN
use a callable macro. So "-width Ar filename" doesn't make sense,
but either "-width Ar" or "-width filename" does, as might something
like "-width xxfilename" for a little extra space.
* There are a lot of needlessly complex hanging tag macros in man4 used
to create simple item lists. Those should be simplified one of these
days before someone copies and edits yet another man4 page.
(see also PR#15179):
When looking up names which directory components (i.e. having slashes,
except when of the form `./name'), FindFile()/DirLookup() first looks
the final filename component in the cache for each directory on the search
path and then proceeds to match the prefixed directory components by
comparing them to the trailing directory components of the the search
path being probed.
This is not correct. When looking for `bar/target' in a path `.../src/foo',
you want it to come up with `.../src/foo/bar/target' (if it exists). There's
no point in comparing the the `bar' prefix on the target to the `foo' suffix
on the search path. Indeed, this will cause a false match if those prefix
and suffix components are actually equal and search path itself also has a
file called `target'. For example, looking for `foo/target' in `.../src/foo'
will spuriously match `.../src/foo/target', not `.../src/foo/foo/target'.
This last bug prompted the change in dir.c, rev 1.27, which happens
to partially workaround it by avoiding the above matching code in the
case of the `curdir' search path entry (at the cost of incurring an
exorbitant amount of cache misses). The situation is unchanged however,
when processing other entries on the search path (e.g. those other than
`dot' and `cur').
Drop the prefix matching code in DirLookup() entirely and use DirFindDot()
and DirLookup() only for names without proper directory components (i.e.
`target' and `./target). Otherwise, non-absolute names are dealt with by
DirLookupSubdir(), while absolute names can be checked for an exact match
of the directory components prefix against the directories on the current
search path. This allows for the use of the file cache to check the
existence of the file and additionally, provides a shortcut out of
Dir_FindFile() if we have the prefix match but not a cache entry (this
is especially beneficial for searches in .CURDIR when it's not equal
to `dot').
the .DOTLAST primitive by a boolean variable with the same name, this whole
mechanism was broken ... it doesn't save much stat calls but it was wrong.
Thanks to Jason Thorpe for the other shadow-variable fixing patches he
made.
foo\:bar:
touch $@
works. Of course I am lazy right now, and I am not removing the '\' escapes
so the output looks fine, but it is functionally correct.
patch makes sure that files included using ".include <bsd.own.mk>" are
really looked for in the system make file directory or in the specified -m
paths instead of allways looking in the other -I and .PATH specified paths
... This speeds up the make a few procents at times for all the system make
files are now found directly instead of searching several paths, saving a
lot of stat() calls.
The number of stat calls is still exorbirant hight though... 910 or so for
making `make' alone ....
the true result works. for some reason the false result doesn't, even
though make -dv shows it being set as the result. (blah!)
- add braces in comments to fix vi showmatch
(XXX - This should be an inclusive, not exclusive, list. Later, this
should be autoconfiscated, but keep the internal implementation, since
src/tools/compat isn't necessarily available for use at make's build time.)
a refCount of 1 to indicate its presence on sufflist. Hence suffixes were
being removed by Suff_EndTransform() and being freed before all the
referential links were removed. This resulted in a malloc warning on 1.5.x:
make in free(): warning: chunk is already free.
This could be reproduced with the following simple Makefile passed to
"make -r":
=====
.SUFFIXES: .l .c
.l.c:
foo:
Instead of skipping the PWD check entirely, add a skip of the PWD override
if MAKEOBJDIR is set and contains a variable transform ($). This has
similar problems to what happens if MAKEOBJDIRPREFIX is set.
file will strip any leading path on the filename when searching ${.CURDIR}.
This bug manifested itself more prominently with xsrc/xfree/xc/lib/GL/GL,
since we now always search ${.CURDIR} as an alternative to ${.OBJDIR}
(PR bin/14499).
breaks too many situations, including MAKEOBJDIR with a :C,foo,bar,
transform in many cases. It's ambiguous and unreliable, as the comment
above that code always indicated. In order to have reliable objdirs, they
need to work the same way Every Time.
(Note that taking this out is not a performance hit; we were already doing
the getcwd() call first. So the getenv("PWD") didn't increase performance.)
* Replace chdir_verify_path() with Main_SetObjdir(), which can be called
externally, and can take a "const char *". (There's a lot of non-const
"char *" passing around in var.c of what should be const strings....)
* Rewrite the initial "find my .OBJDIR" code to make use of the new
function. This still functions as it had in the past, but the comment
above this block was changed to reflect reality: if MAKEOBJDIRPREFIX
or MAKEOBJDIR are set in the environment, then *only that value* is
tried; make does not fall back to obj.MACHINE, obj, and /usr/obj/`pwd`
as it would without these env vars set.
* Add a new special target, .OBJDIR:, which when parsed will cause make to
change to a new object directory and reset .OBJDIR, and PWD in the
environment. This will allow some makefiles (mainly, src/tools)
to override the default objdir semantics in order to add custom logic.
(This splits out the "default system include paths" into its own Lst
variable, and uses it only if sysIncPath is empty. This allows sysIncPath
to be filled in by the Makefile itself.)
make sure that the dir actually exists. In compat mode the corresponding
chdir simply fails and all is well - the issue only arises when playing
tricky games with the value of .CURDIR.
Reviewed: christos
Make sure that each va_start has one and only one matching va_end,
especially in error cases.
If the va_list is used multiple times, do multiple va_starts/va_ends.
If a function gets va_list as argument, don't let it use va_end (since
it's the callers responsibility).
Improved by comments from enami and christos -- thanks!
Heimdal/krb4/KAME changes already fed back, rest to follow.
Inspired by, but not not based on, OpenBSD.
We now just list the names of such variables in .MAKEOVERRIDES.
When we come to export MAKEFLAGS we quote the value of each exported variable
using :Q, using: ${.MAKEOVERRIDES:O:u:@v@$v=${$v:Q}@}
The :O:u suppresses duplicate names.
Also modifed Parse_DoVar to re-export MAKEFLAGS whenever .MAKEOVERRIDES
is assigned to so .MAKEOVERRIDES+= PATH will export PATH=${PATH:Q}
to the environment, while .MAKEOVERRIDES= will disable export of VAR_CMD's.
From Var_Set: We actually want the equivalent of
.MAKEOVERRIDES:= ${.MAKEOVERRIDES:Nname=*} name='val'
clearing the previous value for name is important, since
doing simple duplicate suppression does not handle:
$ make FOO=goo
which then runs a sub-make with FOO=boo
the commands from that sub-make should see just FOO=boo.
via MAKEFLAGS. Instead of appending them directly to .MAKEFLAGS, put
them in .MAKEOVERRIDES (and ensure they are quoted). This is now done
in Var_Set when it exports VAR_CMD's.
Use ExportMAKEFLAGS() to export MAKEFLAGS, using the combined content
of .MAKEFLAGS and .MAKEOVERRIDES (with duplicate supression).
If .MAKEFLAGS is assigned to in a Makefile, ExportMAKEFLAGS is called again.
This allows a line like:
.MAKEOVERRIDES=
to effectively stop the exporting of the command line vars in MAKEFLAGS.
1. make -dx turns on DEBUG_SHELL which causes sh -x to be used where
possible.
2. PrintOnError() is now called when make is stopping due to an error.
This routine reports the curdir and the value of any variables listed
in MAKE_PRINT_VAR_ON_ERROR.
3. Variables set via command line, are propagated to child-makes via
MAKEFLAGS. This behaviour appears to be necessary for POSIX (according
to the GNU folk anyway).
4. Do not reset MAKEFILE when reading ".depend" as this rather eliminates the
usefulness of ${MAKEFILE}.
5. Added ${.newline} as a simple means of being able to include \n in the
result of a :@ loop expansion.
6. Set ${MAKE_VERSION} if defined. Need to come up with a useful value.
Reviewed: christos
Bug reported privately by Nicolas Ollinger <nollinge@ens-lyon.fr>.
You can trigger this bug prior to updating your sources with for example:
% cat > Makefile <<EOF
.DEFAULT:
a
EOF
% make
- install SIGCONT handler which reestablishes SIGTSTP handler and then calls
JobRestartJobs() to restart all the suspended jobs.
- when SIGTSTP is handled, call JobCatchChildren() so *** Suspended messages
get printed before make stops rather than after.
Well, it should say 'make core-dumps on any error on the primary makefile'.
This was a result of the new changes to print the path to the parsed
Makefile... Made the code a lot more conservative, plus prepended <progname>:
to each message.
way to get gettimeofday(), etc. On some systems on which you might want
to host make (e.g. solaris), <time.h> won't get you a struct timeval
definition.
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.
sjg