summaryrefslogtreecommitdiff
path: root/source3/nmbd/nmbd.c
diff options
context:
space:
mode:
authorMartin Pool <mbp@samba.org>2002-03-19 07:26:18 +0000
committerMartin Pool <mbp@samba.org>2002-03-19 07:26:18 +0000
commit1f62003cd771b4620f0f3a5afe9f9999169877f9 (patch)
tree711699023e0a318ab80dba389bdca660cb7b0a50 /source3/nmbd/nmbd.c
parent462402fc098a8d727d92dee8114a3b06d96e3c93 (diff)
downloadsamba-1f62003cd771b4620f0f3a5afe9f9999169877f9.tar.gz
samba-1f62003cd771b4620f0f3a5afe9f9999169877f9.tar.bz2
samba-1f62003cd771b4620f0f3a5afe9f9999169877f9.zip
Change the Makefile to rebuild proto.h as necessary -- note that this
does not imply that all source will be rebuilt when prototypes change, merely that the prototypes will be updated. make proto, clean, delheaders, headers, etc all behave equivalently to before. Intended new behaviour for proto.h, whenever source is being compiled: If proto.h does not exist, it is built. If any source files have changed since proto.h was last checked (.proto.check), then proto.h is checked. If there are no actual changes since last time, its mtime is not changed, but we do remember the time at which it was checked. Whenever we try to build a .o, we need to check the headers are up to date. However, rebuilding the prototypes does not imply rebuilding all object files. Also to allow people to build on machines without Awk, we never try to use it unless a source file has changed. I guess if we wanted, we could have lack of Awk only cause a warning, not failure. The point of all of this is to be easier on people who don't understand or forget to type "make proto", and to reduce the chance of build breakage by having prototypes out of sync. I also rolled back JF's changes to put proto.h into builddir rather than srcdir. There are good arguments in both directions, but since we keep proto.h in CVS, it seems important that the up-to-date copy by in srcdir where it can be checked back in. If people are fussed about having srcdir be readonly you could change this -- but since proto.h is only rebuilt when there are changes, it's not a big deal. I also fixed an apparent race condition in "make headers" that would make it unsafe if you did 'make -j2', and made 'make clean' not kill proto.h, since people may not be able to rebuild it. I reckon there's nothing gnumake-specific here but we shall see. I also have this great idea about rewriting libtool in C++... (This used to be commit 8a61a810e5a29050b0cf242d317c7cc00329517b)
Diffstat (limited to 'source3/nmbd/nmbd.c')
0 files changed, 0 insertions, 0 deletions