. if the build target is invoked again for the install target, the
stack sizes aren't set properly. A workaround is to only build
and not install the servers. (Installing them doesn't really make
sense anyway.)
Thomas Cort [Wed, 29 Jun 2011 14:45:47 +0000 (14:45 +0000)]
opendir(3): on success, don't clobber errno
The opendir(3) function was setting errno to ENOTDIR even
when the directory existed and was opened successfully. This
caused git to falsely detect an error.
This change moves the errno assignment into the failure code
block. It also adds a test to test24 to check for errno
changing when opendir(3) returns success.
Thomas Cort [Mon, 27 Jun 2011 16:01:38 +0000 (16:01 +0000)]
use ${PKGPATH} instead of ${CATEGORIES}/${PKGBASE}
Some packages are in multiple categories (one example is
devel/libgetopt). This broke the IF statement because
${CATEGORIES} got expanded to "cat1 cat2". The proper
variable to use is PKGPATH.
Ben Gras [Thu, 23 Jun 2011 20:43:39 +0000 (22:43 +0200)]
introduce option to time assert()s
. remove a few asserts in the kernel and 64bi library
that are not compatible with the timing code
. change the TIME_BLOCKS code a little to work in-kernel
3 sets of libraries are built now:
. ack: all libraries that ack can compile (/usr/lib/i386/)
. clang+elf: all libraries with minix headers (/usr/lib/)
. clang+elf: all libraries with netbsd headers (/usr/netbsd/)
Once everything can be compiled with netbsd libraries and headers, the
/usr/netbsd hierarchy will be obsolete and its libraries compiled with
netbsd headers will be installed in /usr/lib, and its headers
in /usr/include. (i.e. minix libc and current minix headers set
will be gone.)
To use the NetBSD libc system (libraries + headers) before
it is the default libc, see:
http://wiki.minix3.org/en/DevelopersGuide/UsingNetBSDCode
This wiki page also documents the maintenance of the patch
files of minix-specific changes to imported NetBSD code.
Changes in this commit:
. libsys: Add NBSD compilation and create a safe NBSD-based libc.
. Port rest of libraries (except libddekit) to new header system.
. Enable compilation of libddekit with new headers.
. Enable kernel compilation with new headers.
. Enable drivers compilation with new headers.
. Port legacy commands to new headers and libc.
. Port servers to new headers.
. Add <sys/sigcontext.h> in compat library.
. Remove dependency file in tree.
. Enable compilation of common/lib/libc/atomic in libsys
. Do not generate RCSID strings in libc.
. Temporarily disable zoneinfo as they are incompatible with NetBSD format
. obj-nbsd for .gitignore
. Procfs: use only integer arithmetic. (Antoine Leca)
. Increase ramdisk size to create NBSD-based images.
. Remove INCSYMLINKS handling hack.
. Add nbsd_include/sys/exec_elf.h
. Enable ELF compilation with NBSD libc.
. Add 'make nbsdsrc' in tools to download reference NetBSD sources.
. Automate minix-port.patch creation.
. Avoid using fstavfs() as it is *extremely* slow and unneeded.
. Set err() as PRIVATE to avoid name clash with libc.
. [NBSD] servers/vm: remove compilation warnings.
. u32 is not a long in NBSD headers.
. UPDATING info on netbsd hierarchy
. commands fixes for netbsd libc
Antoine Leca [Mon, 25 Oct 2010 17:03:36 +0000 (18:03 +0100)]
libc!setmode.c: Avoid #if inside macro call
#if inside macro call is undefined behaviour under the C standard
(3.8.3 paragraph 10 for C90, 6.8.10 paragraph 11 for C99).
The same effect can be achieved with a slightly more verbose construct,
putting the whole macro call inside the #ifdef/#else/#endif.
Antoine LECA [Wed, 22 Jun 2011 21:35:26 +0000 (23:35 +0200)]
sys.mk: Improved defaults (sys.mk) for make(1)
. Handle more compiler names, including most cross-compilers.
. Allows to use acd(1) and [whatever-]acc to designate ACK compiler.
. Do not abort (on COMPILER_TYPE not defined) if the compiler name
is not recognized.
sys_umap now supports only:
- looking up the physical address of a virtual address in the address space
of the caller;
- looking up the physical address of a grant for which the caller is the
grantee.
This is enough for nearly all umap users. The new sys_umap_remote supports
lookups in arbitrary address spaces and grants for arbitrary grantees.
Thomas Cort [Thu, 9 Jun 2011 17:12:53 +0000 (17:12 +0000)]
A few aesthetic changes to make the minix port more acceptable
to upstream.
- revert to upstream version of function prototypes for
setting the uid and gid fields of the archive_entry.
- move uid/gid overflow checks into header_common().
- use archive_set_error() instead of fprintf() for getting
error message text back to the main program.
Ben Gras [Tue, 7 Jun 2011 12:48:31 +0000 (14:48 +0200)]
changes for detecting and building for clang/binutils elf
and minor fixes:
. add ack/clean target to lib, 'unify' clean target
. add includes as library dependency
. mk: exclude warning options clang doesn't have in non-gcc
. set -e in lib/*.sh build files
. clang compile error circumvention (disable NOASSERTS for release builds)
Thomas Cort [Tue, 7 Jun 2011 12:46:18 +0000 (12:46 +0000)]
/etc/group: add 'games' group
Several pkgsrc packages under the games category won't install
without a games group. This change adds the group with gid 13,
the default games gid on NetBSD.
Thomas Cort [Mon, 6 Jun 2011 14:31:20 +0000 (14:31 +0000)]
libarchive: fix bad timestamp bug caused by bit shift.
The file timestamps in archives created by libarchive all had
dates in the year 2038. It was caused by a bit shift in
archive_write_set_format_ustar which shifted 1 instead of 1ull.
Tomas Hruby [Fri, 6 May 2011 15:41:14 +0000 (17:41 +0200)]
PCI - do not panic when ACPI cannot map bridges
- when ACPI does not find mappings for pci brdiges, do no panic,
only report a warning and continue to a fallback which uses
only the root bus IRQ routing table. Fail only if that is not
present.
Thomas Veerman [Thu, 21 Apr 2011 13:18:00 +0000 (13:18 +0000)]
Make test40 behave
Make test40 behave better. It should create its own subdirectory to
conduct its tests and should not write to /tmp. Also, the master-slave
terminal pair it tries to open might be in use; it should try to obtain
another pair. These changes allow the test to be run multiple times
simultaneously from different paths (to test select() more intensively).
Kernel: try_async/try_one fixes
- skip processes that are not asynsending to the target
- do not clear whole asynsend table upon IPC permission error
- be more accepting when one table entry is bogus later on
Thomas Veerman [Wed, 13 Apr 2011 13:25:34 +0000 (13:25 +0000)]
Clean up and fix multiple bugs in select:
- Remove redundant code.
- Always wait for the initial reply from an asynchronous select request,
even if the select has been satisfied on another file descriptor or
was canceled due to a serious error.
- Restart asynchronous selects if upon reply from the driver turns out
that there are deferred operations (and do not forget we're still
interested in the results of the deferred operations).
- Do not hang a non-blocking select when another blocking select on
the same filp is still blocking.
- Split blocking operations in read, write, and exceptions (i.e.,
blocking on read does not imply the write will block as well).
- Some loops would iterate over OPEN_MAX file descriptors instead of
the "highest" file descriptor.
- Use proper internal error return values.
- A secondary reply from a synchronous driver is essentially the same
as from an asynchronous driver (the only difference being how the
answer is received). Merge.
- Return proper error code after a driver failure.
- Auto-detect whether a driver is synchronous or asynchronous.
- Remove some code duplication.
- Clean up code (coding style, add missing comments, put all select
related code together).
Server/driver protocols: no longer allow third-party copies.
Before safecopies, the IO_ENDPT and DL_ENDPT message fields were needed
to know which actual process to copy data from/to, as that process may
not always be the caller. Now that we have full safecopy support, these
fields have become useless for that purpose: the owner of the grant is
*always* the caller. Allowing the caller to supply another endpoint is
in fact dangerous, because the callee may then end up using a grant
from a third party. One could call this a variant of the confused
deputy problem.
From now on, safecopy calls should always use the caller's endpoint as
grant owner. This fully obsoletes the DL_ENDPT field in the
inet/ethernet protocol. IO_ENDPT has other uses besides identifying the
grant owner though. This patch renames IO_ENDPT to USER_ENDPT, not only
because that is a more fitting name (it should never be used for I/O
after all), but also in order to intentionally break any old system
source code outside the base system. If this patch breaks your code,
fixing it is fairly simple:
- DL_ENDPT should be replaced with m_source;
- IO_ENDPT should be replaced with m_source when used for safecopies;
- IO_ENDPT should be replaced with USER_ENDPT for any other use, e.g.
when setting REP_ENDPT, matching requests in CANCEL calls, getting
DEV_SELECT flags, and retrieving of the real user process's endpoint
in DEV_OPEN.
The changes in this patch are binary backward compatible.
Thomas Veerman [Fri, 8 Apr 2011 15:23:12 +0000 (15:23 +0000)]
Enable a process to find out what the error code was when delivery of an
asynchronous message resulted in an error.
The model here is that:
- Iff a sender wishes to be notified, the sender MUST check for errors
BEFORE sending another asynchronous message.
The reason is that in order to remember the error code, we can't clean up
the message table and hence we risk running out of table space. This is
less of a problem when the sender enables notifications only for errors.