[odc] Daily src changes for 2005-12-03
ODC
auto at squish.net
Sun Dec 4 07:00:25 GMT 2005
OpenBSD src changes summary for 2005-12-03
==========================================
gnu/usr.bin/binutils gnu/usr.bin/perl
lib/librthread libexec/ftpd
regress/usr.bin sys/arch/hp300/dev
sys/arch/hp300/hp300 sys/arch/luna88k/dev
sys/arch/luna88k/include sys/arch/luna88k/luna88k
sys/arch/m88k/conf sys/arch/m88k/include
sys/arch/m88k/m88k sys/arch/mac68k/mac68k
sys/arch/mvme68k/dev sys/arch/mvme88k/dev
sys/arch/mvme88k/include sys/arch/mvme88k/mvme88k
sys/dev/eisa sys/dev/ic
sys/dev/isa sys/dev/pci
sys/dev/usb sys/kern
sys/sys usr.bin/bc
usr.bin/cvs usr.bin/rcs
usr.bin/xlint
== gnu =============================================================== 01/06 ==
http://www.openbsd.org/cgi-bin/cvsweb/src/gnu
usr.bin/binutils
~ gdb/Makefile.in ~ gdb/defs.h
~ gdb/solib.c ~ gdb/solib.h
~ gdb/utils.c ~ gdb/doc/observer.texi
+ gdb/bsd-uthread.c + gdb/bsd-uthread.h
> Threads support. (kettenis@)
~ gdb/i386obsd-tdep.c ~ gdb/config/i386/obsd.mt
~ gdb/config/i386/obsd64.mt
> i386 threads support. (kettenis@)
usr.bin/perl
~ sv.c
> Use PERL_INT_MAX instead of INT_MAX in previous commit for consistency
> with the rest of perl. On OpenBSD PERL_INT_MAX is the same as INT_MAX.
> From rse at engelschall.com.illert@)
== lib =============================================================== 02/06 ==
http://www.openbsd.org/cgi-bin/cvsweb/src/lib
librthread
+ Makefile + rthread.c
+ rthread.h + rthread_attr.c
+ rthread_sched.c + rthread_sig.c
+ rthread_sync.c + rthread_tls.c
+ arch/amd64/_atomic_lock.c + arch/amd64/rfork_thread.S
+ arch/i386/_atomic_lock.c + arch/i386/rfork_thread.S
> add userland thread library. incomplete, but functional (tedu@)
~ rthread.c ~ rthread_sched.c
> syscall is actually sched_yield now, as millert suggested (tedu@)
~ Makefile
> rcs for the makefile (tedu@)
== libexec =========================================================== 03/06 ==
http://www.openbsd.org/cgi-bin/cvsweb/src/libexec
ftpd
~ ftpd.c
> remove shadowing variables; ok pval (deraadt@)
== regress =========================================================== 04/06 ==
http://www.openbsd.org/cgi-bin/cvsweb/src/regress
usr.bin
~ xlint/Makefile ~ xlint/test-11.c
+ xlint/test-11.c.exp
> Add a couple more __attribute__ test cases to test 11, and hook up 11
> now that lint deals with attributes. (cloder@)
== sys =============================================================== 05/06 ==
http://www.openbsd.org/cgi-bin/cvsweb/src/sys
arch/hp300/dev
~ mb89352.c
> The first thing done when XXX_scsi_cmd() returns TRY_AGAIN_LATER is
> to set xs->error to XS_BUSY. So it is pointless and misleading to
> set xs->error to XS_TIMEOUT, XS_DRIVER_STUFFUP or XS_BUSY just before
> returning TRY_AGAIN_LATER.
> No functional change.
> ok miod@ (krw@)
arch/hp300/hp300
~ clock.c
> splimp -> splvm
> ok miod@ (brad@)
arch/luna88k/dev
~ mb89352.c
> The first thing done when XXX_scsi_cmd() returns TRY_AGAIN_LATER is
> to set xs->error to XS_BUSY. So it is pointless and misleading to
> set xs->error to XS_TIMEOUT, XS_DRIVER_STUFFUP or XS_BUSY just before
> returning TRY_AGAIN_LATER.
> No functional change.
> ok miod@ (krw@)
arch/luna88k/include
~ locore.h
> Switch m88k ports to __HAVE_CPUINFO. Current cpu pointer is held in SR0
> on all running processors.
> Tested aoyama@ and I (miod@)
~ locore.h
> Turn read_processor_identification_register() into a simple macro with a
> much, much, much shorter name. It is only used to print cpu revision
> anyway... (miod@)
+ mutex.h
> Fast __HAVE_MUTEX implementation for m88k platforms. (miod@)
+ lock.h
> Implement __cpu_simple_lock_t for m88k. (miod@)
arch/luna88k/luna88k
~ eh.S ~ locore.S
~ m8820x.c ~ machdep.c
> Switch m88k ports to __HAVE_CPUINFO. Current cpu pointer is held in SR0
> on all running processors.
> Tested aoyama@ and I (miod@)
~ locore.S
> intstack does not need to be visible from C code anymore. (miod@)
arch/m88k/conf
~ files.m88k
> Switch m88k ports to __HAVE_CPUINFO. Current cpu pointer is held in SR0
> on all running processors.
> Tested aoyama@ and I (miod@)
~ files.m88k
> Fast __HAVE_MUTEX implementation for m88k platforms. (miod@)
arch/m88k/include
~ asm.h ~ cmmu.h
~ cpu.h ~ pcb.h
~ types.h
> Switch m88k ports to __HAVE_CPUINFO. Current cpu pointer is held in SR0
> on all running processors.
> Tested aoyama@ and I (miod@)
~ asm_macro.h
> Turn read_processor_identification_register() into a simple macro with a
> much, much, much shorter name. It is only used to print cpu revision
> anyway... (miod@)
~ types.h + mutex.h
> Fast __HAVE_MUTEX implementation for m88k platforms. (miod@)
~ intr.h + lock.h
> Implement __cpu_simple_lock_t for m88k. (miod@)
~ cmmu.h ~ pmap.h
> Replace simplelocks with __cpu_simple_locks for cmmu and pmap locking,
> for the MULTIPROCESSOR case. (miod@)
arch/m88k/m88k
- cmmu.c ~ eh_common.S
~ genassym.cf ~ m88k_machdep.c
~ pmap.c ~ process.S
~ subr.S ~ trap.c
~ vm_machdep.c
> Switch m88k ports to __HAVE_CPUINFO. Current cpu pointer is held in SR0
> on all running processors.
> Tested aoyama@ and I (miod@)
~ m8820x_machdep.c ~ subr.S
> Turn read_processor_identification_register() into a simple macro with a
> much, much, much shorter name. It is only used to print cpu revision
> anyway... (miod@)
~ eh_common.S
> intstack does not need to be visible from C code anymore. (miod@)
~ pmap.c
> Dissociate pmap locks and spl usage; code will now do explicit spl
> processing
> in addition to pmap locking; no functional change. (miod@)
~ genassym.cf + mutex.S
> Fast __HAVE_MUTEX implementation for m88k platforms. (miod@)
~ m8820x_machdep.c ~ m88k_machdep.c
~ pmap.c
> Replace simplelocks with __cpu_simple_locks for cmmu and pmap locking,
> for the MULTIPROCESSOR case. (miod@)
arch/mac68k/mac68k
~ intr.c
> splimp -> splhigh
> From NetBSD
> ok miod@ (brad@)
arch/mvme68k/dev
~ sbic.c ~ ssh.c
~ vs.c
> The first thing done when XXX_scsi_cmd() returns TRY_AGAIN_LATER is
> to set xs->error to XS_BUSY. So it is pointless and misleading to
> set xs->error to XS_TIMEOUT, XS_DRIVER_STUFFUP or XS_BUSY just before
> returning TRY_AGAIN_LATER.
> No functional change.
> ok miod@ (krw@)
arch/mvme88k/dev
~ vs.c
> The first thing done when XXX_scsi_cmd() returns TRY_AGAIN_LATER is
> to set xs->error to XS_BUSY. So it is pointless and misleading to
> set xs->error to XS_TIMEOUT, XS_DRIVER_STUFFUP or XS_BUSY just before
> returning TRY_AGAIN_LATER.
> No functional change.
> ok miod@ (krw@)
arch/mvme88k/include
~ locore.h
> Switch m88k ports to __HAVE_CPUINFO. Current cpu pointer is held in SR0
> on all running processors.
> Tested aoyama@ and I (miod@)
~ locore.h
> Turn read_processor_identification_register() into a simple macro with a
> much, much, much shorter name. It is only used to print cpu revision
> anyway... (miod@)
+ mutex.h
> Fast __HAVE_MUTEX implementation for m88k platforms. (miod@)
+ lock.h
> Implement __cpu_simple_lock_t for m88k. (miod@)
arch/mvme88k/mvme88k
~ eh.S ~ locore.S
~ m188_machdep.c ~ m88110.c
~ m8820x.c ~ machdep.c
> Switch m88k ports to __HAVE_CPUINFO. Current cpu pointer is held in SR0
> on all running processors.
> Tested aoyama@ and I (miod@)
~ m88110.c
> Turn read_processor_identification_register() into a simple macro with a
> much, much, much shorter name. It is only used to print cpu revision
> anyway... (miod@)
~ locore.S
> intstack does not need to be visible from C code anymore. (miod@)
~ m88110.c
> Remove unused debug material. (miod@)
~ m88110.c
> Replace simplelocks with __cpu_simple_locks for cmmu and pmap locking,
> for the MULTIPROCESSOR case. (miod@)
dev/eisa
~ aha1742.c
> The first thing done when XXX_scsi_cmd() returns TRY_AGAIN_LATER is
> to set xs->error to XS_BUSY. So it is pointless and misleading to
> set xs->error to XS_TIMEOUT, XS_DRIVER_STUFFUP or XS_BUSY just before
> returning TRY_AGAIN_LATER.
> No functional change. (krw@)
dev/ic
~ mpt.c ~ mpt.h
~ mpt_openbsd.c ~ mpt_openbsd.h
> Remove some typedef struct goo. (marco@)
~ gdt_common.c
> Don't retry unsupported SCSI commands by returning TRY_AGAIN_LATER.
> They aren't going to be magically supported the next time. Elminates a
> long delay at boot time, since TRY_AGAIN_LATER code was recently fixed
> to delay a second before retrying and MODE SENSE commands were
> reclassified as unsupported.
> ok tdeval@ marco@ (krw@)
~ aac.c ~ adv.c
~ adw.c ~ aic6360.c
~ ami.c ~ bha.c
~ cac.c ~ dpt.c
~ gdt_common.c ~ iha.c
~ isp_openbsd.c ~ osiop.c
~ siop.c ~ trm.c
~ twe.c ~ uha.c
> The first thing done when XXX_scsi_cmd() returns TRY_AGAIN_LATER is
> to set xs->error to XS_BUSY. So it is pointless and misleading to
> set xs->error to XS_TIMEOUT, XS_DRIVER_STUFFUP or XS_BUSY just before
> returning TRY_AGAIN_LATER.
> No functional change. (krw@)
~ if_wi.c ~ if_wi_hostap.c
> splimp -> splnet
> ok marco@ (brad@)
dev/isa
~ wds.c ~ seagate.c
~ aha.c
> The first thing done when XXX_scsi_cmd() returns TRY_AGAIN_LATER is
> to set xs->error to XS_BUSY. So it is pointless and misleading to
> set xs->error to XS_TIMEOUT, XS_DRIVER_STUFFUP or XS_BUSY just before
> returning TRY_AGAIN_LATER.
> No functional change. (krw@)
dev/pci
~ pcidevs.h ~ pcidevs_data.h
> Regen (pedro@)
~ pcidevs
> Add ATI IXP200 Modem, from Bruno Oliveira Silvestre, okay deraadt@ (pedro@)
dev/usb
~ ehci.c ~ ohci.c
~ uhci.c
> Allow strings descriptor 0 to be fetched. It's the default language index.
> From augustss NetBSD
> ok dlg@ (brad@)
~ if_wi_usb.c
> splimp -> splnet
> ok marco@ (brad@)
kern
~ init_main.c ~ init_sysent.c
~ kern_exec.c ~ kern_exit.c
~ kern_fork.c ~ kern_kthread.c
~ kern_proc.c ~ kern_prot.c
~ kern_sig.c ~ kern_synch.c
~ syscalls.c ~ syscalls.master
> kernel support for threaded processes (rthreads).
> uses rfork(RFTHREAD) to create threads, which are presently processes
> that are a little more tightly bound together. several new syscalls
> added to facilitate a userland thread library.
> all conditional on RTHREADS, currently disabled.
> ok deraadt (tedu@)
~ init_sysent.c ~ syscalls.c
> rebuild (tedu@)
sys
~ param.h ~ proc.h
~ syscall.h ~ syscallargs.h
> kernel support for threaded processes (rthreads).
> uses rfork(RFTHREAD) to create threads, which are presently processes
> that are a little more tightly bound together. several new syscalls
> added to facilitate a userland thread library.
> all conditional on RTHREADS, currently disabled.
> ok deraadt (tedu@)
~ syscall.h ~ syscallargs.h
> rebuild (tedu@)
== usr.bin =========================================================== 06/06 ==
http://www.openbsd.org/cgi-bin/cvsweb/src/usr.bin
bc
~ scan.l
> ARGSUSED signal handler (deraadt@)
cvs
~ checkout.c ~ cmd.c
~ cvs.h ~ diff3.c
~ entries.c ~ file.c
~ file.h ~ import.c
~ rcs.c ~ rcs.h
~ remove.c ~ resp.c
~ update.c ~ util.c
> add very basic support for the following stuff:
> - checkout in local mode (example: /cvs)
> - update in local and server mode (example: /cvs and user at hostvs)
> - import in local and server mode (example: /cvs and user at hostvs)
> what remains to be done:
> - not all options are supported yet, and update cannot pick up newly
> added files yet. these things are pending and will be commited
> real soon.
> - checkout only works locally right now.
> - fix rcs parsing code so that we don't fucking hog 100% cpu
> on really BIG BIG BIG ass trees.
> mainly tested by pedro@ and myself, thanks a lot pedro!
> "go for it" niallo@ (joris@)
~ rcs.c
> usr.bin/rcs stuff doesn't care about cvs_checkout_rev(); (joris@)
~ file.c ~ file.h
> remove unused and useless functions, they just get in the way; (joris@)
~ util.c
> nuke unused vars; (joris@)
~ update.c
> respect Q flag in output; (joris@)
~ cvs.h ~ entries.c
> - teach opencvs about Entries.Log and what to do with it
> it it exists in the CVS admin dir.
> - when writing the Entries file write it to Entries.Backup first and
> then rename it to Entries when the writing was successfull. (joris@)
~ cvs.c ~ cvs.h
~ file.c ~ rcs.c
~ util.c
> fix date handling in opencvs, this was broken since the very start.
> now, when you checkout a tree with gnu cvs, opencvs no longer sees all
> the files as modified, and visa versa.
> this actually makes gnu cvs and opencvs dance together in working copies.
> (joris@)
~ rcs.c ~ rcs.h
~ update.c
> todd@ reminds me:
> If you 'vi file.c' and change stuff, then 'make depend && make' you'll
> have a file.o file most definately more recent than the latest committed
> version of file.c.
> Then, if you rm file.c, and cvs update file.c, if 'file.c' is set to the
> timestamp of the last committed version, a subsequent 'make' will not
> rebuild the .o file.
> to fix this we simply use the current timestamp when updating files.
> thanks todd@! (joris@)
~ req.c
> strip any trailing slashes in the received Directory
> request, to make sure we are compatible with gnu; (joris@)
rcs
~ ci.c
> - add support for -t[description] and -t-[file containing description]
> (niallo@)
~ ci.1
> - document `-t' option. (niallo@)
~ ci.c
> - whoops, its -t-string or -tfile, not the other way around
> - sync checkin_usage() with reality (niallo@)
xlint
~ lint1/cgram.y ~ lint1/lint1.h
~ lint1/scan.l
> Lint can now parse every variation of gcc's __attribute__ that I could find
> either in our tree or in the gcc docs. See regression test 11 for examples
> of this. Right now, our cdefs.h actually defines __attribute__(x) to blank
> when lint runs. We can change this whenever we choose. (cloder@)
~ lint1/lint1.h ~ lint1/scan.l
> Fix enum types, thanks lint (cloder@)
===============================================================================
More information about the odc
mailing list