• make fast character classes even faster by removing the C_SUBOP2 hack
in favour of a separate seldom-used ksh_issubop2 macro (which also
makes ctype() side-effect-safe) which is a slower class (no change there)
• optimise cases of ksh_isalphx followed by a ksh_isalnux loop
(used parsing variable names)
• remove a misleading comment in initctypes() about \0 from pdksh
• rename C_ALPHA to C_ALPHX to make it more clear the underscore is included
• sprinkle a few ord() in there
• add new ksh_isalpha() which tests for [A-Za-z] (slow character class)
• there is no '_:\' drive on OS/2 (which inspired the whole changeset)
related to that:
• while AT&T ksh may do it, POSIX says nothing about allowing
declaration commands only without vars and redirections, and
“without vars” especially seems against which commands they are
• fix relationship between forwarders and real declaration commands
• clean up c_builtin vs shcomexec mess
Also, re-run “make repool” with a fixed src/scripts/stringpool.sh,v 1.3
• if HAVE_STRING_POOLING is set to 1
• if HAVE_STRING_POOLING is set to 2 and not GCC < 4 is used
• if HAVE_STRING_POOLING is not set to 0 and LLVM or GCC >= 4 is used
Closes: LP#1580348
prodded by izabera and carstenh; resolution is:
• you can’t trim a vector in mksh, still (consider ${@:-1})
• future POSIX will require non-empty “word” for most “op”s
• dissolve in order of standard → extension
• dissolve to prefer “op” over “prefix” where still necessary, mostly
• ord() new, From: Daniel Richard G. <skunk@iSKUNK.ORG>
‣ used in some places
• (c - '0') → ksh_numdig(c) # may take *x++ argument
• (c - 'A') → ksh_numuc(c) # may NOT take *x+= argument
‣ idem for ksh_numlc(c) and 'a'
‣ these need changing for EBCDIC
‣ add testsuite for this
• use macros more, they exist already often
• use digits_lc[foo] instead of ('0' + foo), especially for letters
• caught another ksh_eq case…
• also caught a maybe-UB overflow check, but we don’t have TIME_T_MAX ☹
so we mitigate a bit (in var.c mostly) and tweak another type already, and
add some checks (mksh_{,u}ari_t must fit into {,unsigned }long) and print
line numbers with %lu already
bug initially found by Pawel Wylecial (LP#1440685)
additional bug found and suggested fix by enh (elliott hughes)
This commit also renames struct ioword.flag to ioflag to disambiguate
it from other members named “flag”, changes it to an unsigned type,
and packs ioflag and unit into shorts each, to make the struct smaller
(aligned even: 16 bytes on 32-bit systems) and reviews some of the
code involved in fd handling, though there wasn’t much to be found.
tg