2. do not use _CTYPE_NUM_CHARS macro to read data from LC_CTYPE(old BSDCTYPE style) database.
because 1<<CHAR_BIT is MD, so i added MI macro _CTYPE_CACHE_SIZE(1<<8).
3. remove _NB_CACHED_RUNE macro, use _CTYPE_CACHE_SIZE instead.
They implement a space efficent write-once database with fast access
path. Switch the services(5) database to use cdb. The size of the
database file decreases from 2.1MB disk space to 307KB. Access
performance is about the same if setservent(0) is used and about an
order of magnitude faster otherwise. services_mkdb defaults to the new
format, but can optionally create the old db(3) format as well for
statically linked legacy applications.
libc implementation (such as *BSD and glibc2).
2. don't typedef void * wc{type,trans}_t, suggested by soda@-san.
it may pass through compiler type check, it's harmful.
so i introduce dummy struct __tag_wc{type,trans}_t(iconv_t already does).
no ABI change was made.
instead of depending on the existance of struct winsize. Technically I should
bump the library version or version the symbol, but nothing seems to use this
outside the library!
in struct nlist, since it's accessed differently depending on
whether we do a.out or not. Use this macro as appropriate in the
nlist* functions.
Also replace some includes of <a.out.h> with <nlist.h>, to fix build
problem for mips.
the latter pulls FD_SET etc and it makes wine-1.1.27's winsock vs unixsocket
namespace dance unhappy. for more information, grep
Include_winsock_h_before_stdlib_h_or_use_the_MSVCRT_library
in their source.
gnulib, the implementation goes back to the AMD Software Optimizer
guide. A number of platforms will want to replace the C version with
assembler code using native instructions.
The algorithm used is the Jenkins hash. The name (mi_vector_hash)
reflects the nature of the hash function.
Add glue for libc ATF tests and include a test case to make sure that
(mis)alignment and endianess are handled correctly.
Bump libc minor to 169.
There was no room to allocate 16 bytes for _rnd in the struct in the _LP64
case, so we dynamically allocate now. Also put the _rnd field last, so that
we are bitwise compatible too. Now of course we've run out of space, so
next time will need to perform more evil tricks to fit.
tnozaki