cvs commit: src/sys/conf files options src/sys/dev/video/bktr ioctl_bt848.h msp34xx.c CHANGELOG.TXT bktr_audio.c bktr_audio.h bktr_card.c bktr_card.h bktr_core.c bktr_core.h bktr_i2c.c bktr_i2c.h bktr_mem.c bktr_mem.h bktr_os.c bktr_os.h bktr_reg.h ...

Matthew Dillon dillon at apollo.backplane.com
Mon May 17 00:31:26 PDT 2004


:But this doesn't work if you only want to recrate /usr/include e.g. via the
:recommented "rm -R /usr/include; mkdir /usr/include; make installincludes".
:
:sigh. Why does buildworld use symlinks for the temporary include space anyway?
:The space usage (10 MB) is negligible and it would save us from such head
:aches.
:
:Joerg

    It depends whether you want to have to rebuild the whole world whenever a
    sys/ file changes (for example) or whether you incremental builds to work
    with a larger set of changes.  I'd prefer to keep the current flexible
    incremental build capability for things like sys file changes.

    I'd like to see a requirement that the buildworld sequence not access 
    include files through those alternate symlinks (most of which are network
    related).  Limiting their creation to just the installed non-symlinked version
    of the include files and not generating the extra symlinks at all for the
    buildworld's temporarily generated includes would be a reasonable solution.
    That would also probably simplify the ioctl scanning code, which I think 
    has some makefile mess to remove duplicate subdirectories in order to
    avoid duplicate declaration errors from the compiler.

    I'd prefer to minimize world breakage as much as possible, though.  Any
    changes with that level of sophistication need to be very well tested before
    they are committed.

					-Matt
					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>





More information about the Commits mailing list