move /etc/{,suid_}profile to /system/etc/ for Android (but do not make
the location of /etc configurable); rewrite manpage section about
/etc/{,suid_}profile, .profile, .mkshrc
• while here, reformat 'struct tbl' comment-wise and placement-wise
and drop the Tflag typedef
• while here, write regression test for the "global" built-in, which
does what typeset is supposed to do except that it doubles as "local"
• IBM XL C: display version better (tested on V7.0 by cnuke@)
• do not 'IFS=: read nr name', Cygwin 1.7 dash fails it
• disable cd-pe, glob-range-3 on Cygwin (the former cannot
succeed because the mv fails, the latter fails from 1.7 on)
• mark heredoc-tmpfile-8 as need-pass: no
• apply __attribute__ only to a function prototype, not to
the body (even if static), since xlC fails that
• bump version to R40 (beta)
Testsuite:
• add new need-pass: {yes|no} attribute, default yes
• exit with 1 if a need-pass test failed unexpectedly
idea by Kacper Kornet <draenog@pld-linux.org>
• mark utf8bom-2 as need-pass: no
Infrstructure:
• add housekeeping function for making a tty raw
• switch functions with unused results to void
• struct op: u.charflag contains last char of ;; in TPAT
• var.c:arraysearch is now a global function
Language:
• add ;& (fall through) and ;| (examine next) delimiters
in addition to ;; (end case) as zsh extensions, because
POSIX standardised on ;& already
• add -A (read into array), -N (read exactly n bytes),
-n (read up to n bytes), -t (timeout) flags for read
from ksh93
• allow read -N -1 or -n -1 to slurp the entire input
• add -a (read into array the input characters) extension
specific to mksh to read, idea by David Korn
• add -e (exit with error if PWD was not set correctly
after a physical cd) to cd builtin, mandated by next
POSIX, and change error codes accordingly
Rewrites:
• full rewrite of read builtin and its manpage section
• add regression tetss for most of the new functionality
• duplicate hexdump demo tests for use of read -a
• use read -raN-1 in dot.mkshrc to get NUL safe base64,
DJB cdb hash and Jenkins one-at-a-time hash functions
• don’t leak memory parsing possible I/O redirection tokens
• get rid of volatile by using more const (also helps codegen, methinks)
• support empty here document markers (mksh extension)
• pimp the manpage
• ensure that bool/true/false are cpp macros, overriding any pre-defined
• document the requirement that tobool(x) must map any-type 'x' into bool
• document the requirement that a bool must only be true or false, and
that it (tobool() rather) must have an identity mapping to 'short'
• possibly fix ksh_func for/and fpFUNCTf – maybe spotted by cnuke@
a bit more with POSIX and the other shells
I considered http://austingroupbugs.net/view.php?id=253 but the use
of bi_errorf() is interesting, especially as it’s often enough a
noreturn function, and funnily enough, 'cd -P /foo' returns 0 while
'chdir -P /foo' fails (so idk where to put -e)…
⇒ consider merging simplify_path()
• move funcs.c:c_cd() to misc.c
• make misc.c:make_path() static, c_cd() is its only user
⇒ mark as obsolete
• tweak misc.c:set_current_wd() to drop ksh_get_wd() argument
should be no code change, but the entire path stuff is a mess…
so expect actual implementation changes or even rewrites shortly
and switches to the TARGET_OS=Linux
• introduce android as regression test suite category
• add an android specific standard alias
• clean up redundant ‘-o sh’ arg in a few checks
UTF-8 BOM instead (UTFMODE has a separate value now for activated
during BOM skipping)
• parsing a COMSUB now skips UTF-8 BOM, too, but only temporarily
(token stream, lexer output / parser input), EOS terminated, let
SASPAREN use the same lexing as SBASE (e.g. COMSUB recursively)
• make wdstrip recursive
• fix processing of COMSUB in wdstrip
⇒ pass comsub-1 test
• expose another debugging function
• use shf_putc (macro), shf_putchar (function) ipv tputc
• replace shf_putchar(x,y) calls for side-effect-less x with shf_putc
• plug another bug in the tree code – '\' → "\\" (backslashes must be
escaped inside double quotes, too)
• adjust testsuite (and, I _had_ wondered…)
EOF) # works again now, plugging a regression
• rewrite the here document parsing code to be *much* more efficient
(and a bit more readable too!) using goto, while here (no kidding)
① currently: ((cond) ? true : false) but (!!(cond)) and casting to bool,
the latter only if stdbool.h, would also work – which performs best on
(and across) all supported systems?
instead, but the parser for the so-called “backticks” (U+0060) still emits
plaintext COMSUB wdstrings, and the evaluation code emits plaintext if the
code is not run (‘-n’ option), so it’s not worth the effort and memory ma-
nagement issues, even though it _would_ optimise the most common case…
Bump version numbers, sync regression tests; add one testcase from the old
webpages too. Sync manpage, this now works, but keep the workaround in, as
“portability issue” with slightly changed wording.
Also, /bin/sleep must be used in one manpage example if sleep is built in.
• functions called by mksh’s grouping memory allocator
• functions called by mksh code itself
the latter may be changed to call the internal grouping allocator,
if a porter so desires (but if this were recommended, the code in
question would already do so, so…)
• PIPESTATUS now supported (like bash 2) whose last member
may actually differ from $? since the latter may not be the
result of a pipeline partial command
• add regression tests, documentation, etc.
• in interactive mode, always look up {LC_{ALL,CTYPE},LANG} environment
variables if setlocale/nl_langinfo(CODESET) doesn’t suffice
• add the ability to call any builtin (some don't make sense or wouldn't
work) directly by analysing argv[0]
• for direct builtin calls, the {LC_{ALL,CTYPE},LANG} environment
variables determine utf8-mode, even if MKSH_ASSUME_UTF8 was set
• when called as builtin, echo behaves POSIXish
• add domainname as alias for true on MirBSD only, to be able to link it
• sync mksh Makefiles with Build.sh output
• adjust manpage wrt release plans
• link some things to mksh now that we have callable builtins:
bin/echo bin/kill bin/pwd bin/sleep (exact matches)
bin/test bin/[ (were scripts before)
bin/domainname=usr/bin/true usr/bin/false (move to /bin/ now)
• drop linked utilities and, except for echo and kill, their manpages
• adjust instbin and link a few more there as well
a mirtoconf check, would’ve been a real problem on an LP64 platform
• sh.h: work around a bad interaction between -Wformat on gcc and manual
string pooling for T_synerr, which is used in place of a format string
in some places
– possible integer overflows in memory allocation, mostly
‣ multiplication: all are checked now
‣ addition: reviewed them, most were “proven” or guessed to be
“almost” impossible to run over (e.g. when we have a string
whose length is taken it is assumed that the length will be
more than only a few bytes below SIZE_MAX, since code and
stack have to fit); some are checked now (e.g. when one of
the summands is an off_t); most of the unchecked ones are
annotated now
⇒ cost (MirBSD/i386 static): +76 .text
⇒ cost (Debian sid/i386): +779 .text -4 .data
– on Linux targets, setuid() setresuid() setresgid() can fail
with EAGAIN; check for that and, if so, warn once and retry
infinitely (other targets to be added later once we know that
they are “insane”)
⇒ cost (Debian sid/i386): +192 .text (includes .rodata)
• setmode.c: Do overflow checking for realloc() too; switch back
from calloc() to a checked malloc() for simplification while there
• define -DIN_MKSH and let setmode.c look a tad nicer while here
we don’t get SIGWINCH when the window size changes during the runtime of
that, so, the signal is only usable reliably during editing in the shell
and we re-check the window size before each interactive edit line again
a string buffer whose window size is currently 32 (initial), your data
is 96 bytes, this routine used to resize the buffer to 64, append your
first 64 bytes to it (no matter if there's already something in it)
and then writes the remaining bytes to stdio fd instead of the string…
if it doesn’t SIGABRT before
discovered by wbx@ – thanks – bug inherited from pdksh 5.2.14 (AD 1999)
• deactivate %a and %A since our libc doesn’t have it
• rewrite the mksh integration code to use shf instead of stdio, removing
floating point support always in the process, as shf doesn’t support it
⇒ saves 11114 (6706 text, 168 data, 4240 bss) with dietlibc on Debian
• fix -Wall -Wextra -Wformat -Wstrict-aliasing=2 for gcc (Debian 4.4.4-7)
• fix these and -Wc++-compat for gcc version 4.6.0 20100711 (experimental)
[trunk revision 162057] (Debian 20100711-1) except:
– a few enum warnings that relate to eglibc’s {g,s}etrlimit() functions
taking an enum instead of an int because they’re too stupid to adhere
to POSIX interfaces they design by themselves
– all “request for implicit conversion” involving a "void *" on one side
• tweak the manual page somewhat more
• avoid calling realloc twice in sequence, since the final
size is known at the first call already
• do not lstat(2) the same path twice in the Hurd codepath
concurrently accessing the same $HISTFILE be more synchronised with
each other: empty lines (just pressing Return) and duplicates (that
are split and written twice by the lines loaded from $HISTFILE in
the meantime); requested by Maximilian “mxey” Gaß in #!/bin/mksh
some idiotic terminal emulators and/or people seem to use the es-
cape codes normally denoting Alt-Arrowkey instead so let's simply
bind them to the vt_hack as well... (untested)
• merge the rest of branch tg-wcswidth-behaviour
• enhance test cases for wcswidth-like behaviour
• switch hash table collision resolution algorithm to Python’s as announced
• bump vsn
│remember to restore errno (ie. stop someone from making a mistake later)
│ok guenther
check.t, sh.h: bump vsn
I wonder though why errno must be restored even if nothing was
called after reading it… moid?
which, in its latest sid incarnation, even received mksh's ability
to produce ${!foo[*]} array keys, wow!)
* plug a memory leak while here (ATEMP only, but still)
I read, IIRC in the Cederqvist, that 'cvs tag' sets a sticky tag onto
the cwd… it doesn’t, apparently. (I actually like it better this way,
but one needs to know!)
others (colon and equals sign need to be simply escaped, while dollar
sign and accent gravis need double escaping like opening square brak-
ket did back then); add = to C_QUOTE to simplify (doesn't break any-
thing) and sort these strings asciibetically while here
• use a combination of the one-at-a-time hash and an LCG for handling
the $RANDOM special if !HAVE_ARC4RANDOM instead of rand(3)/srand(3)
and get rid of time(3) usage to reduce import footprint
• raise entropy state (mostly in the !HAVE_ARC4RANDOM case though…)
• simplify handling of the $RANDOM_SPECIAL generally
• tweak hash() to save a temp var for non-optimising compilers
• some int → mksh_ari_t and other type fixes
• general tweaking of code and comments
just a "somewhat more POSIX" but also a "/bin/sh legacy kludge" mode
* consistently capitalise POSIX and SUSv3/SUSv4 (same as AT&T ksh) and
Bourne shell
to it are now either arc4random or rand/srand, but srand retains the old
state; set +o arc4random is no longer possible, but if it's there we use
arc4random(3), if not, we use rand(3) for $RANDOM reads; optimise special
variable handling too and fix a few consts and other minor things
MKSH_S_EDIT for small (Emacs) editing mode, MKSH_S_FEAT for all the dis-
abled language features), which can be set to 0 despite MKSH_SMALL being
defined to re-enable the Vi command line editing mode (which I wouldn't,
but fits into the general mastermind scheme)
some GNU bash extensions (suggested by cnuke@) and bind macros
* make the random cache more efficient (and the code potentially
smaller, although we have a new implementation of the oaat hash
function, alongside the old one, now) and pushb only if needed
(i.e. state has changed or user has set $RANDOM, but not onfork)
• shell flags are now handled in one single place (sh_flags.h)
• sync comments (between enum and array) and manpage with reality
• FMONITOR is now no longer needed for Hartz IV shells
integers in addition to my 「1#a」 (or 「1#…」), which also allows for
finer end-of-character checking. Note that this is locale-dependent in
ksh93, set ±U dependent in mksh, and mksh’s OPTU-16 encoding is used.
libc function realpath(3) which may not be available on the target
system; compile the realpath builtin unconditionally
looks fine to me, but review is appreciated; this is (very) lightly
based upon MirBSD libc’s realpath(3) and pdksh’s get_phys_path()
it with the array index; var.c says that
│ 1244 /* The table entry is always [0] */
so that we can have a special flag and a union which stores hval for
the table index, the array index otherwise (coïncidentally *hint hint*
they have the same size)
return information needed to do a real ktremove instead of the pseudo
ktdelete operation which merely unsets the DEFINED flag to mark it as
eligible for texpand garbage collection (even worse, !DEFINED entries
are still counted)
much better avalanche and no known funnels
• improve comments
• fix some types (uint32_t for hash, size_t for sizes)
• optimise ktsort()
no functional change, I think
Build.sh but use 'if defined(PRECOND) && !defined(TOBEDEFINED)'if possible
* for all of the source code, drop annotations "imake style" (if we check
for specific OSes, bad, instead of using mirtoconf checks proper) and
"conditions correct?" (if I'm not entirely sure if that #if catches all
cases and no false positives) where I can see it by grepping immediately
* bump mksh patchlevel
* refresh Makefiles
more from cid 1004A2D72DD5A4E4B4F tried to be fixed in 1004A300A72701188E3
but I’d appreciate someone who actually uses Vi Mode to test it:
Revision 1.26: [7]download - view: [8]text, [9]markup, [10]annotated - [11]select for diffs
Mon Jun 29 22:50:19 2009 UTC (5 days, 14 hours ago) by martynas
Branches: [12]MAIN
CVS tags: [13]OPENBSD_4_6_BASE, [14]OPENBSD_4_6, [15]HEAD
Diff to: previous 1.25: [16]preferred, [17]coloured
Changes since revision 1.25: +10 -5 lines
make VSEARCH werase act like regular werase after the last change.
vi back-words and emacs kill-region are not completely the same.
ok merdely@, millert@. "Get it in" Darrin Chandler
starting with an ‘!’ exclamation mark at the beginning of a com-
mand (PS1 not PS2), shall have the same effect as the predefined
“r” alias, to be compatible with csh and GNU bash’s “!string” to
«Execute last used command starting with string» – documentation
and feature request provided by wbx@ (Waldemar Brodkorb)
and it actually REDUCES code size to allow it as well; mention
in the manpage that it’s merely unportable (and of course exe-
cution time differs); sync clog
• expose “#ifdef MKSH_MIDNIGHTBSD01ASH_COMPAT” just in case they decide to
require it and show it in the ksh version automatically
• sync the use of non-ASCII characters over files (unification)
fix the regression test’s results while here, which have been
broken since cid 10049D9BE5254CE65B8
• get rid of separate copyright file which was intended for De-
bian; track down commits in all files of oksh-mirbsd and mksh
to get correct copyright years per-file, as is BSD custom
${foo:1:2} operates on characters ipv bytes – which means:
‣ set +U: octets
‣ set -U: MirOS OPTU-8 characters
for consistency I also adapted ${#stringname} to deliver the
length in characters ipv bytes; more may follow; for example
I’d like a way to expose the string width.
you can already get the MirOS OPTU-16 of a character in the
WTF-8 (「set -U」) mode with something like
│ typeset -Uui16 -Z7 x=1#${stringname:position:1}
which will correctly use the PUA EF80‥EFFF mapping for octets.
due to this being an incompatible change, bump to R38
also change the unicode-hexdump sample regression test and
add two news for ${x:1:2} and ${#x} checks in A/W mode ☺
struct env (other structures defined have no "foreign type with pos-
sible alignment constraints" members) and take care of it while dea-
ling in a struct env instance
gcc, SUNWcc, pcc, llvm-gcc, clang, etc. all didn't say a thing!
now compiles warning-free (testsuite pass) on ULTRIX 4.5 (1986),
and OSF/1 X2.0-8 (testsuite norun: perl missing) has only the usual
bitchings about "volatile sig_atomic_t" because the latter part is
already volatile, but otherwise warning-free compile, works fine
pass "xerrok" status across the execution call stack to more closely
match what both POSIX and [18]ksh.1 already describe in regards to set
-e/errexit's behavior in determining when to exit from nonzero return
values.
specifically, the truth values tested as operands to &&' and ||', as
well as the resulting compound expression itself, along with the truth
value resulting from a negated command (i.e. a pipeline prefixed !'),
should not make the shell exit when -e is in effect.
issue reported by matthieu.
testing matthieu, naddy.
ok miod (earlier version), otto.
man page ok jmc.
similarily simple one from scratch, which however performs
better than espie's with AFREE_DEBUG enabled which took away
the benefit of the double-linked-list approach
all of (core) mksh is now MirOS licenced
H4sIAAAAAAACA31TTUvDQBC951c804DtIQ0iiFhSPHiwiCDqzUpIuhO6JN2ETSS16n93Np+GFiGE
zM6bN2/fTPQOro5xENYuEVLzBz9eaF6RlWupSpyHaXqzVuvyljbbDCRV0UR7WeLiHEtT4D2GCcUy
pdNFh4pkg4wGZCV8Z6opTPOw3GI+swoq4ZIVZxqkSv0JqRAiWkBkmAJNegHIGG9vcAWcSnjM6dTo
uby8vsL7+wLllhQQKKr4SoFv/8nbXE5pQcdpk4nlwH6AE7zer17uVs/BmHVftnUdxYh0yHbV3ghY
9wh2rQlBkdMm8PmIe57Bjfs7dTW9X/OoEG6lwzwnzYKA7+8jHpsJTuFN4+IjKhKZM6dpZ0xuQ2Nz
7bGB1F74zlebmzTm/PS+OB0GZz4HLcPgz+BeDxyS9S7YU6e3qZaQk5gZgRHvQtI59Eet7QwBWFnL
O/F+ej9Fpmg0ug7OHrnqHy2+7y8x6DGEG3E0Az41RuPlYfXEEx0psrl8PAUebzf6ZklabgD8V3SS
Z7Vo6xfOuQS6gQMAAA==
「mksh -o posix z」 failed in that it continues; 「mksh z」 correctly aborts
let’s see what the obsd people have to say herefore
H4sIAAAAAAACAz1PywrDIBA8m68YgoT20EN7TMixX1F6yGNFIWhRSw2h/95VmuDB2XmsY2oDRVyo
Si1N2uHlMzQK0b8JHaImy4RQwxIITYPiUs4xqcyeGgdfrfumtWX5dMbGSHy0WQgP1PJa49lhdpkV
ynncYSxkOjjxd/V83dmcX9sQtFGRi0zORmO5042HbwnMzlIBqa9lAmfLVBIZ7XqpKBPDnDu+WXpi
wMhnOgQXYvUD+oKHAhUBAAA=
XXX OpenBSD has something different which may DTST or even DTRT (not break
XXX our make(1) wrt <bsd.subdir.mk>), check that
on Debian Lenny/amd64 (XXX need more verification; this
can be used for 64 bit arithmetics later too)
PPID, PGRP, RANDOM, USER_ID are now unsigned by default
on the screen is not enough for two columns, just output the text line by
line, instead of trying to format it; gets rid of superfluous empty lines
if we did not even have space for one column on the screen (x_cols)
noticed by Gábor Gergely in irc, thanks!
was hard to type and hard to fix, galloc is also hard to fix, and some
things I learned will probably improve things more but make me use the
original form as base (especially for space savings)
* let sizeofN die though, remove even more casts
* optimise, polish
* regen Makefiles
* sprinkle a few /* CONSTCOND */ while here
* syn.c: fix bashiop-4 regression test; failed due to me using a
simple string when a wdstring was expected, sorry; the new code
assumes file descriptors take up a maximum of two characters
encountered. However, when reading end of input, the source type is set
to SEOF while popping, whereas the recursion check code only checks for
an SALIAS type.
Fix: add a new SF_HASALIAS flag; change u.tblp from being valid if type
is SALIAS to being valid if SF_HASALIAS is set; set SF_HASALIAS for the
created SALIAS sources; set SF_HASALIAS and u.tblp when creating SALIAS
whose next is SEOF on the SEOF source as well.
Reported by Michael Hlavinka as Redhat Bug #474115
allocator using malloc and free, with mmap malloc and omalloc in mind,
not counterfeiting its security measures such as guard pages, and having
some of our own, e.g. XOR random cookies, optional mprotect, etc.
zero cost (for we have arc4random())
$ (CCC_LD=mgcc CC=ccc sh Build.sh -r && ./test.sh -v) 2>&1 | tee log
Total failed: 2 (as expected)
Total passed: 278
Just the result is huge, and we could of course build to intermediate
byte code to optimise globally…
please pcc, prompted for by Anders “ragge” Magnusson, problem spotted
originally by Adam “replaced” Hoka
⇒ rewrote x_bs2() and utf_backch() into a combined x_bs3() function,
since these are never used in any other way
• whitespace cleanup, while here
“-sh” if -DMKSH_BINSHREDUCED was passed during compilation, for example
for Debian, but d̲e̲f̲i̲n̲i̲t̲i̲v̲e̲l̲y̲ n̲̲o̲̲t̲̲ for MirBSD™
• split up regression test to force this behaviour
• remove the gunk from our MirBSD™ startup scripts again
• mention arc4random.c changes on website, sync clog, warn packagers
abortion (^G – ^C is SIGINT and doesn’t work like this, but
that’s actually good IMO)
prompted by enquiry about the Emacs editing mode by <smultron:#MidnightBSD>
and .data instead of another initialisation; this was prompted by a bug
in scan-build (the value can never be NULL, but it doesn’t realise it),
although this doesn’t fix it, but less stack usage is always good
fool the compiler into not doing static bounds checking when we do
one-past-the-array-boundary pointer assignments for cases where the
only accesses are like (*--pointer); bump version
in a somewhat hackish way, and it’s still quite different from zsh,
but probably closer to a desired functionality
XXX this makes state by abusing 「modified」 and 「xmp」 (“the mark”).
‣ only if !MKSH_SMALL
‣ add appropriate regression test
• if FPOSIX is set, do not close fds > 2 on exec, Debian #499139
• add appropriate regression tests for keeping fds private or not
• fix vi mode (which, however, is officially orphaned) multi-line $PS1 by
using a similar algorithm for prompt skipping as emacs mode (changing
the meaning of prompt_trunc variable and using prompt_redraw, just even
more efficiently than vi mode); reported by asarch via IRC
• fix multi-line prompts if last line is “too large” by using emacs mode
algorithm of just internally appending a newline, while here ☺ this even
saves us having to re-add the prompt_skip variable…
WARNING: this is only barely tested, as almost nobody ever uses vi mode
⇒ test yourself, there may be bugs (e.g. off-by-ones); already known is
that the vi input line editing mode is NOT multibyte safe…
since mksh(1) did go into an infinite loop if that fails first
bug spotted, initial patch and help drafting a test case
From: Decklin Foster <decklin@red-bean.com>
note there are more instances of unlink(2) and others (like chmod(2), as
spotted by flawfinder) which aren’t checked… but at least the other case
of unlink(2) use in histrap.c doesn’t cause any trouble (I think)
and make it fit into mksh’s model (also gives us a couple of things
GNU bash doesn’t have
• add regression tests for all of these
Lukas “smultron” Upton from MidnightBSD spotted a script with /bin/sh
shebang invalidly using “&>” in some Apple backup toolkit, 10x
XXX why fds are limited to one digit?
‣ macro afreechk() is superfluous
• get rid of macro afreechv() by re-doing the “don’t leak that much” code
• some KNF (mostly, whitespace and 80c) while here
* initialise the integers PPID, OPTIND, RANDOM, SECONDS, and TMOUT to base-10
* bring back PGRP as base-10 integer to the process group via getpgrp(2)
* initialise USER_ID as base-10 integer to the effective user id as retrieved
from geteuid(2) = $(id -u)
* use $USER_ID in dot.mkshrc instead of spawning an id(1) process
-> dot.mkshrc,v 1.34 now requires mksh R34
* convert more int to bool where appropriate
* remove dead code - getpgrp(2) cannot fail
* sync manual page to reality
* bump to mksh R34(beta) - feature freeze
XXX check if our_pgrp in jobs.c is still really needed, the setpgid call
XXX probably just makes us our own pgrp leader, and we might have to use
XXX and update kshpgrp accordingly - need feedback/help here but I think
XXX this simplification should be possible if I grok the code correctly.
etc/profile:
* adjust to $USER_ID changes in mksh (speed-up here, too)
mksh.hts:
* sync changelog
• more int → bool
• more regression tests: check if the utf8-hack flag is really disabled
at non-interactive startup, enabled at interactive startup, if the
current locale is a UTF-8 one
• make the mksh-local multibyte handling functions globally accessible,
change their names, syntax and semantics a little (XXX more work needed)
• optimise
• utf_wctomb: src → dst, as we’re writing to that char array (pasto?)
• edit.c:x_e_getmbc(): if the second byte of a 2- or 3-byte multibyte
sequence is invalid utf-8, ungetc it (not possible for the 3rd byte yet)
• edit.c:x_zotc3(): easier (and faster) handling of UTF-8
• implement, document and test for base-1 numbers: they just get the
ASCII (8-bit) or Unicode (UTF-8) value of the octet(s) after the ‘1#’,
or do the same as print \x## or \u#### (depending on the utf8-hack flag),
plus support the PUA assignment of EF80‥EFFF for the MirBSD encoding “hack”
(print doesn’t, as it has \x## and \u#### to distinguish, but we cannot use
base-0 numbers which I had planned to use for raw octets first, as they are
used internally): http://thread.gmane.org/gmane.os.miros.general/7938
• as an application example, add a hexdumper to the regression tests ☺
(3 weeks, 5 days ago) by millert
Make ulimit able to get and set multiple limits in a single invocation
like bash and zsh do. Requested by espie@, OK deraadt@
• apply diff from mirbsdksh-1.11:
#ifdef DUP2_BROKEN
/* Ultrix systems like to preserve the close-on-exec flag */
‣ XXX we do #ifdef __ultrix here (imake-style) instead of mirtoconfing it
(but does anyone know of any other OS with the same problem? plus we’d
see it as we now know the symptoms)
• remove ultrix Build.hs warn=' but might work…' in the hope it DOES
• split Xinit into XinitN and Xinit macro, the former
not initialising the “xp” argument of the latter,
and use this to get rid of two variables that are
only assigned but never referenced (gcc doesn’t see
this, but MIPSpro and IIRC SUNWcc do)
• re-indent while here
• bump patchlevel
‣ I/O redirection seems broken:
$ (date; date >/dev/null; date) | wc -l
1 (expected: 2)
‣ other than that: working fine
‣ -YBSD (default) and -YSYSTEM_FIVE don’t work, just -YPOSIX, somehow
• Fix $(…) to `…` for OSF/1 V2.0 /bin/sh
‣ this compiler is FUBAR though:
$ cat >t.c
main() { return (foo()); }
$ cc t.c
ld:
Unresolved :
foo
$ echo $?
0
$ ls -l a.out
-rwxr-xr-x 1 mirbsd users 10835 Jul 21 17:12 a.out
‣ it seems to have ucode, but man is not installed
• new mirtoconf check: mkstemp(3)
• if !HAVE_MKSTEMP (Ultrix), use tempnam(3)
• only use printf(1) if it exists (it doesn’t on Ultrix)
• a few more signals
• add S_ISLNK if the OS doesn’t define it
• add strcasecmp(3) proto for Ultrix (it _is_ in <portability.h>, but
only for -YBSD I think)
• fgrep(1) on Ultrix doesn’t do “-e ① -e ②”
10x DEChengst:#UnixNL for giving access
| csh -cf '/command/svscanboot &'
and
| /usr/mpkg/bin/pgrphack /usr/mpkg/bin/svscanboot &
can now be replaced with
| /bin/mksh -T- /usr/mpkg/bin/svscanboot
| OSF1 rubbereendje.dechengst.nl V5.1 2650 alpha
with the vendor compiler:
| Compaq C V6.5-011 on HP Tru64 UNIX V5.1B (Rev. 2650)
| Compiler Driver V6.5-003 (sys) cc Driver
• the platform’s sig_t is incompatible too (simplify check)
• no compile warnings at all
• results in:
$ size mksh
| text data bss dec hex
| 327680 16384 17808 361872 58590
$ file mksh
| mksh: COFF format alpha dynamically linked, demand paged executable or object module not stripped - version 3.13-14
$ ldd mksh
|
| Main => mksh
| libc.so => /usr/shlib/libc.so
$ ls -l mksh
| -rwxr-xr-x 1 mirbsd users 395200 Mar 5 19:18 mksh
• minor testsuite issues:
FAIL ./check.t:regression-13
unexpected stderr - got too much output
wanted nothing
got:
Successful
cat: output error
⇒ probably harmless
• works like a charm!
todo tomorrow:
• test case (compare with e.g. GNU bash)
• manpage
• version bump
sqchar is a bit ugly, but \/ must be preserved, as we don’t get wdencoded
strings later on in the process (eval.c CSUBST) and I didn’t want to have
an implementation like ${foo: 2: 3} this time
this change broke abortion on failure to read input, was not
needed for gcc warnings and is the fault of Intel’s compiler
this should fix the other busy-loop problem occuring only on
GNU/Linux so far – 10x spaetzle@freewrt.o for pointing me to
the problem; reproduced on my work craptop
from netbsd via oksh
we had the NULL pointer deref already fixed
• avoid a bogus not-setting the return value of edit.c:x_file_glob()
introduced by the above change in oksh
• escape ? as well (but not ] because that’s wrong)
reminded by cbiere@netbsd via oksh
• Unsetting a non-existent variable is not an error. See
http://www.opengroup.org/onlinepubs/009695399/utilities/unset.html
report from Arkadiusz Miskiewicz; fixed based on
http://cvs.pld-linux.org diff via oksh but modified slightly
• Be more smart waiting for input for non-interactive scripts. Fix
based on a diff from debian: see their bug#296446 (via oksh)
modified slightly
this also fixes cnuke@’s “mksh busy loop” problem, for which I never
received a bug report, but the Debian bug page contains a set of two
scripts to reproduce this before (and no longer after) this commit
• some KNF
• bump version
In contrast to AT&T ksh93, its semantics are like GNU bash in that it ap-
pends the current working directory to the search path; it is implemented
as a shell alias instead of enhancing funcs.c:shbuiltins[] like in ksh93.
delivered to the process
• regression test by Clint Pachl, verified against Debian ksh93 by me
• place to fix it in the code discovered by Otto Moerbeek
namely Dr. Robert “Pfeffer” Arnold (in this case, in FreeWRT), make
a half-completed attempt at implementing ${foo:2:3} substring evals
(of course, negatives can't work right now and that the numbers are
in face expressions is something I only read later too – this is to
be revisited later, but it's already late)
don't depend on this behaviour yet though
if someone wants to add more regression tests, feel free to…
is a compromise anyway; these lunox people will have to live with that, too
many existing korn shell alike scripts depend on it even if not on the full
korn shell syntax availability (note: this doesn't mean using these in some
script with #!/bin/sh is ok)
| tglaser@hephaistos:~ $ /usr/local/intel/cc/9.1.042/bin/icc -V
| Intel(R) C Compiler for 32-bit applications, Version 9.1 Build 20060706Z Package ID: l_cc_c_9.1.042
this one is muuuuch too verbose IMHO, _and_ it #defines __GNUC__ (eek)
car only slower, yes this is possible, and the resulting binary passes the
testsuite just fine), the definition of __RCSID() in <sys/cdefs.h> expands
to something with __attribute__((used)), which triggers a warning, because
__attribute__ in general is supported but the used attribute isn't. Thusly
always use our own strings and get rid of the MULTI_RCSID test (introduced
because __RCSID() on Darwin is inferiour).
Maybe we should fix <sys/cdefs.h> too? #ifdef __SUNPRO_C helps here.
it was #if solaris'd in R28, and the system I have access
on does declare it now (still Solaris 8)
this can be put back if anyone complains, of course.
it's wrong to use strchr(s, 0) to look for the NUL byte, because in some
environments it apparently might return NULL
use new macro strnul = s+strlen(s) instead (not side-effect safe tho)
convert options() prototype to unsigned (size_t, in fact), and make an
explicitly casted (size_t)-1 the error return code, modelled after what
is often used in Unix libraries
Analysis:
internal_errorf(int, fmt, ...) was only a __dead function if the int argument
was non-0, which the Prevent probably was unable to follow. Change all uses of
internal_errorf(0, fmt, ...) to internal_warningf(fmt, ...); change the pro-
totype of internal_errorf() to internal_errorf(fmt, ...) and all remaining
uses remove the non-0 int argument; add __dead to internal_errorf() proto;
flesh out guts of internal_errorf() and internal_warningf() into a new local
function for optimisation purposes.
Some whitespace cleanup and dead code removal (return after internal_errorf(1))
XXX u_int32_t is not ISO C99, but seems to work well enough
XXX if it fails anywhere, we'll see in the build logs
XXX apple doesn't have the standard uint32_t and API doesn't specify it
• sys_siglist_defn: rename to sys_siglist_decl as we're really checking
for the declaration in the headers; change wording to “check if … does
not need to be declared” since we don't need to declare if we don't use
• scan for arc4random, arc4random_push, confstr declarations too
• sh.h: confstr declaration is no longer #ifdef __sun__; sort
do not start with a "#!shell" sequence” part of mksh(1) true again; this
was probably lost in mksh R21 or so when I decided/saw that our kernel
always parses shebang lines (code written myself, not taken from pdksh
again or MirBSD kernel, but verified against both)
• bom+shebang execution now works, no need setting EXECSHELL in the test
• bump version
given to execute, standard input (interactive or not), via -c command line
argument, or after “eval”, but not for $(…) comsubs, at the beginning of a
subsequent line, or within a line, etc.); regression test for it
idea during my “week off” (despite the pain), bsiegert@ thinks it's good –
and utf-8 capable tools ought to be able to do this anyway
and have it return an API-correct const char *
• enhance and stylify comments
• a little KNF and simplifications
• #ifdef DEBUG: replace strchr and strstr with ucstrchr and ucstrstr
that take and return a non-const char *, and fix the violations
• new cstrchr, cstrstr (take and give const char *)
• new vstrchr, vstrstr (take const or not, give boolean value)
• new afreechk(x) = afreechv(x,x) = if (x1) afree(x2, ATEMP)
• new ksh_isdash(str) = (str != NULL) && !strcmp(str, "-")
• replace the only use of strrchr with inlined code to shrink
• minor man page fixes
• Minix 3 signames are autogenerated with gcc
• rename strlfun.c to strlcpy.c since we don't do strlcat(3) anyway,
only strlcpy(3), and shorten it
• dot.mkshrc: move MKSH=… down to the export line
to not disturb the PS1 visual impression ☺
• dot.mkshrc: Lstripcom(): optimise
• bump version
¹) side effect from creating API-correct cstrchr, cstrstr, etc.
uses goto so it must be better ☻
tested on mirbsd-current via both Makefile and Build.sh
sequence is eaten before the command is called; cought by <TGEN>
(Thomas E. Spanjaard) via IRC
fix is to tabcomplete a newline to singlequote+newline+singlequote
* bump version
side (I don't have capacities to scan 71711 files in MirOS for a
standards-compliance whack)
* copyright: mention CVS changelogs as place of contributors' name
and credit information, too
* check.t, sh.h: bump
where we had 'noreturn' etc. but no '__noreturn__')
* Scan for __attribute__((bounded)) and __attribute__((used))
if we have __attribute__((noreturn))
* To be able to scan if certain attributes give warnings,
scan for -Werror with a simple programme which hopefully triggers none
* Convert __attribute__((unused)) to __unused, noreturn -> __dead
* Unify other attributes
* Clean up typography a little more
* 'sigseen' in Build.sh goes away
* Signal name existence is checked in this order:
have our own¹ -> sys_signame[] -> _sys_signame[] -> build our own²
* Signal description existence is checked in this order:
sys_siglist[] -> _sys_siglist[] -> strsignal() -> NULL³
¹ Predefined list of items, for operating systems where we
cannot build² them, i.e. Plan 9 and Minix 3 (e.g. no $CPP -dD)
² The usual cpp(1) stuff
³ Changed later, see below
* Make $CPP test dependent on $NEED_MKSH_SIGNAME (others can
be added here, this is not absolute)
* Make signal name list generation² dependent on $NEED_MKSH_SIGNAME
* Fix check if the generation worked
* Guarantee that sigtraps[*].name and sigtraps[*].mess are valid
C strings; this makes the code shorter *and* removes a few pos-
sible nil pointer dereferences
* Embed autoconf'd usages of sys_sig* / strsignal / mksh_sigpairs
into inittraps()
* Check for each signal 0<=i<=NSIG that
name is not NULL or "" -> replace with ("%d", i)
mess is not NULL or "" -> replace with ("Signal %d", i)
name does not start (case-insensitive) with "SIG" -> name += 3
* In gettrap(), fix check if signal name starts, case-sensitive
or case-insensitive, depending on need, with "SIG" (bug from millert@)
Other changes:
* Build.sh: ac_test[n]() are documented
* Build.sh: ac_test[n]() can have negative prereqs as well now
* Build.sh: use <<-'EOF' consistently
* bump patchlevel to today
to strcasestr, it was used in a wrong way (reverse logic error in
checking its return value), turning to mis-detection of UTF-8 locale.
* sh.h, check.t: bump version
* copyright: bump year
"set +o emacs-usemeta" and "set -o vi-show8" which are always on now,
since we have proper internationalisation (i.e. utf-8) support, and
assume the user either has a 'C' locale and can't enter 8-bit chars,
his terminal is 8bit-transparent, or he has a 'UTF-8' locale.
twkm (from #ksh on freenode), that $RANDOM is always an unsigned
15-bit decimal integer.
(RANDOM << 15 | RANDOM) thusly yields 30 bit, which is still more
than 36^5, so we can use it on the baselife CD to speed things up
main.c: In function 'main':
main.c:208: warning: cast discards qualifiers from pointer target type
main.c:329: warning: cast discards qualifiers from pointer target type
no warnings at autoconf time left either; will take care of these two later
(might revisit changes from this commit), maybe change declararion for the
builtins to have their argv[] be const strings, and go through strict type
and qualifier checking again. this'll further improve stability.
XXX these changes might have introduced (more?) memory leaks,
XXX someone who knows about these tools should verify with
XXX automatic memory usage analysers (valgrind?)
still passes testsuite
XXX one of these uses a gcc extension, ok for now tho
* don't include <ctype.h> any more at all
* don't try nl_langinfo in small mode, just check locale
saves 171 .text, 4 .data, 256 .bss, 1 import
* edit.c: remove debug stuff again; next time better use shl.c functions ;)
* sh.h: add format attributes to a few shf functions
* histrap.c, var.c: fix format string mistakes
* main.c, sh.h: error_prefix and warningf take bool not int
* misc.c: make chvt() stuff use shf_* functions
* misc.c: rewrite the TIOCSTTY stuff to be better integrated in mksh,
since it originally was an external patch
* misc.c: chvt() no longer fails if e.g. chown fails due to e.g. R/O / fs
* var.c: fix typeset padding for right-justified zero-filled
tg
ahoka