| Age | Commit message (Collapse) | Author |
|---|
|
Preparations to fix the engine:
- new function has_been_built(gn), that tells you what's the status of
a given node. Allows us to run Suff_FindDeps later, by updating the number
of unmade children correctly.
- take out the code that handles shell expansions in an expand_children*
set of functions, called by Suff_FindDeps, among others. These must be
called early in the engine to avoid creating bogus nodes.
Engine fixes:
- take the predecessor/successor special handling out, deal with it in
separate functions.
- don't count nodes. Explicitly track them all in a hash table (better
way to deal with non-built issues).
- don't run Suff_FindDeps at start, but just before building an actual node.
This allows make to find all dependencies correctly, as in groff.
Pfiou! now it works.
|
|
This is a really big step towards getting parallel make to work.
Note that this is not yet complete. There are still a few `details' to
fix before this works 100%. Specifically: sequential make (compat) and
parallel make don't use the same engine, and the parallel engine still
has a few limitations. For instance, some known issues:
- parallel make does not deal with .phony targets correctly all the time.
- some errors are deadly in parallel make mode.
- parallel make NEEDS way more sturdy correspondance of file system paths
and target names, since it often needs to match dependencies to targets
before the corresponding files exist.
- some local variables like $* get set in a bogus way in some cases.
- suffix handling has issues, especially related to the NULL suffix.
So, if you find stuff that does NOT yet work with parallel make, don't go
blindly try to fix the Makefile. It's very likely you might have stumbled
into a make bug. (unless you really, really, understand Makefiles, DON'T
GO CHANGING THEM YET).
Tested by lots of people, thanks go to miod@, and robert@ among other people.
Quick summary of what this does:
- remove `saving commands' extension (it's not really usable, nor used)
- move compat job runner and parallel interrupt handling into engine.c
- tweak the code so that both compat and parallel mode use the same job runner
and the same interrupt handling. Remove the other one.
- optimize job runner so that, in parallel mode, the last command does not
fork if we can avoid it (as it's already running in a sub shell).
- scrape all the code that dealt with creating shell scripts from commands.
- scrape all the code that dealt with recognizing special sequences in
command output to print/not print output.
- fix the parallel job pipe to not keep around file descriptors that are not
needed.
- replace the parallel job buffering with a nicer one, that deals with
non-blocking descriptors to try to agregate as much output from one job in
one go (greed) to unconfuse the users.
- create two pipes per job, so that stdout and stderr stay separate.
- make job token printing a debug option.
- always use the parallel job-runner to `execute' commands, even if we just
print them out.
- store list of errors encountered during parallel make running, and print them
on exit, so that we know what went wrong.
- add a dirty hack to targ.c to deal with paths produced by gccmakedep.
|
|
|
|
Store special targets in target hash, and use them for the parsing.
Use OP_DUMMY flag to mark targets that don't really exist yet, such
as interrupt and default nodes.
Also, .PATHxxx is special in suffixes.
Small tweaks to compat.c, so that run_commands does more stuff after
the fork() (and thus no need to free things).
Remove distinction between local and global jobs.
|
|
to target nodes at all (reduces modules inter-dependencies)
|
|
- cut up those huge include files into separate interfaces for all modules.
Put the interface documentation there, and not with the implementation.
- light-weight includes for needed concrete types (lst_t.h, timestamp_t.h).
- cut out some more logically separate parts: cmd_exec, varname, parsevar,
timestamp.
- put all error handling functions together, so that we will be able to
clean them up.
- more systematic naming: functioni to handle interval, function to handle
string.
- put the init/end code apart to minimize coupling.
- kill weird types like ReturnStatus and Boolean. Use standard bool (with a
fallback for non-iso systems)
- better interface documentation for lots of subsystems.
As a result, make compilation goes somewhat faster (5%, even considering
the largish BSD copyrights to read). The corresponding preprocessed
source goes down from 1,5M to 1M.
A few minor code changes as well: Parse_DoVar is no longer destructive.
Parse_IsVar functionality is folded into Parse_DoVar (as it knows what an
assignment is), a few more interval handling functions. Avoid calling
XXX_End when they do nothing, just #define XXX_End to nothing.
Parse_DoVar is slightly more general: it will handle compound assignments
as long as they make sense, e.g., VAR +!= cmd
will work. As a side effect, VAR++=value now triggers an error
(two + in assignment).
- this stuff doesn't occur in portable Makefiles.
- writing VAR++ = value or VAR+ +=value disambiguates it.
- this is a good thing, it uncovered a bug in bsd.port.mk.
Tested by naddy@. Okayed millert@. I'll handle the fallback if there is
any. This went through a full make build anyways, including isakmpd
(without mickey's custom binutils, as he didn't see fit to share it with me).
|
