these commits move all path handling into module_do_load() from
kobj_load_file(). This way the final path used to load a module
is available for loading <module>.plist, which will store parameters
for a module. The end goal of this project is good support for
MODULAR device drivers.
XXX This should be done as part of an overall plan to support
power management and device detachment. However, in order to do
that, I would first have to invent sbus_intr_disestablish(). This
is being done at this time in order to aid in the effort to eliminate
shutdownhook_establish().
This was based on the sys/arch/sparc64/fd.c change. Thanks to jdc@
for testing this version.
This was writing time_t to disk. Worse, the time_t member was at an
unaligned offset in the structure in question, so after the time_t
change the structure layout depended on compiler-inserted structure
padding. This gives three legacy formats: one with 32-bit time, one
with 64-bit time, and one with 64-bit time and 4 bytes of structure
padding.
And of course the file didn't have a header or version coding or
anything.
The new code writes a structure of well-defined size that should not
receive unexpected padding, and gives the file a header and version
number. It reads that format and any of the three legacy formats,
figuring out which one it's dealing with by inspecting the file
contents. For good measure, it also now handles opposite-endian files,
doesn't bail out unceremoniously unless necessary, and won't croak if
the file is corrupt and e.g. contains unterminated strings.
(Was it worth going to this length? Maybe not. But it didn't seem
right to just leave it, and it's not clear where to stop halfway.)
+ get rid of some magic constants
+ revamped regression test script to count number of tests passed
+ made checkhash array in ops_seckey_t dynamic, rather than statically
allocated
+ made mdc array dynamic, and added a length field to mdc for future use
+ revamped usage message to match reality
+ made portable version again for the autoconfed package sources
+ add separate netpgpdigest.h file so that separate digest sizes can be
used without having to include "packet.h" in everything
Some object file diffs, but they are harmless. (Mostly they seem to
come from internal counters in gcc... and in one case the order of two
instructions was harmlessly swapped, which is odd and annoying.)
neither has done anything in a long long time. Add previously
undocumented -s/-f (slow vs. fast) options, although I question the
utility thereof and suspect they should just be removed.
Bump date (first time since 1993)
numbers out.
This changes the "tournament codes"; that is, the same code will give
you a different game now from what it used to. (This is because the
codes are basically random seeds.) I really really doubt anyone cares
about this, especially since the tournament feature appears to be
undocumented.
Default to "fast mode" as ~nobody has a 300 baud terminal any more.
("Fast mode" apparently controls whether short-range scans are printed
by default at certain times.)
- Avoid atomics in more places.
- Remove the per-descriptor mutex, and just use filedesc_t::fd_lock.
It was only being used to synchronize close, and in any case we needed
to take fd_lock to free the descriptor slot.
- Optimize certain paths for the <NDFDFILE case.
- Sprinkle more comments and assertions.
- Cache more stuff in filedesc_t.
- Fix numerous minor bugs spotted along the way.
- Restructure how the open files array is maintained, for clarity and so
that we can eliminate the membar_consumer() call in fd_getfile(). This is
mostly syntactic sugar; the main functional change is that fd_nfiles now
lives alongside the open file array.
Some measurements with libmicro:
- simple file syscalls are like close() are between 1 to 10% faster.
- some nice improvements, e.g. poll(1000) which is ~50% faster.
