merged:
• new regression tests
• check.pl (tests/th) better tmpfile handling
• exec.c 1.50: POSIX specifies that for an AND/OR list,
only the last command's exit status matters for "set -e"
• ksh.1 1.147: document the above
• eval.c 1.39: “Make $(< /nonexistent) have the same behaviour
as $(cat /nonexistent) wrt. errors (do not unwind and do not
treat this as fatal if set -e is used).”
‣ additionally make shf_open() return errno and actually show
the error message from the system
• regression-39 test: remove the “maybe” marker
‣ but decide on correct POSIX behaviour
already been fixed in mksh:
• check.pl (tests/th) exit 1 if tests fail
not merged:
• main.c 1.52: run traps in unwind() before exiting;
I’m pretty sure this is already working as-should in mksh
• eval.c 1.38: “Commands executed via `foo` or $( bar ) should
not inherit "set -e" status.” As discussed in IRC, this is
just plainly WRONG.
• sig_t detection was a bit insane, it is a function-pointer type after all
• fix uninitialised variable in c_select which led to mistakenly accepting
invalid (nōn-numeric) input and acting, randomly, upon it
• keep SIGCHLD blocked in child after forking longer, for job list manip
• block SIGCHLD ifdef DEBUG_LEAKS to not run job foo during/after afreeall
• fix annoying ISO C90 vs. C99 (un)signed constant warning
command; setting one still unsets the other at first)
• Change subst_exstat to be conformant unless -o sh is set and -o posix isn’t
• In lksh, make subst_exstat (newly) conformant if -o posix
• New MKSH_BINSHPOSIX to accompany MKSH_BINSHREDUCED
• Sync lksh manpage precisely
• correct order of built-in commands; use POSIX special versus “all others”
plus “keeps assignments” as distinction, no longer play POSIX regular vs.
others game; sync manpage
• fix LP#1156707: map (( internally to “let]” which is no valid function
name and so can’t be overridden but is unlikely to be used otherwhere
and not strictly permitted (by POSIX) anyway
• we do not need -Wno-overflow any more, either
• bump to R45
that get used, plus one for the realpath-1 regression test; also make
sys_siglist_decl detection nicer and poison strerror() with non-const
return value ifdef DEBUG, make it always const
• tty_fd is now never closed
• new tty_hasstate tracks tty_state (cf. thread around
http://article.gmane.org/gmane.os.miros.mksh/79 and PLD bug)
• as users requested, importing COLUMNS or LINES from the environment
now removes its special-ness as does unsetting it
• otherwise, setting COLUMNS or LINES is honoured until the next SIGWINCH
arrives or change_winsz is otherwise run (e.g. before displaying the
prompt in the interactive command line editing modes)
• SIGWINCH is now honoured before each reading of $COLUMNS and $LINES too
• change the Uhr to match – it no longer calls stty(1) ☺
• all writers of exstat ensure the value is in [0; 0xFF]
• all readers of exstat AND it with 0xFF (not strictly needed thus)
• trap_exstat is “safe”, i.e. always either -1 or [0; 0xFF]
several conditions are met as outlined below; for more background, refer to
http://www.FreeBSD.org/cgi/query-pr.cgi?pr=113860
We don’t yet optimise 「% sh -c '{ echo a; sleep 10;}&'; sleep 1; ps T」 so
the FreeBSD® sh approach cannot work for us, but scanning the “sh -c” argu‐
ment for disallowed characters and, if not, setting a flag that enables the
shell to exec the tree when parsed as TCOM *and not c_trap()* was possible.
Disallowed characters are currently C_QUOTE except space, that is:
Tab Newline " # $ & ' ( ) * ; < = > ? [ \ ] ` |
This should catch all cases of magic, variables, subshells, pipelines, etc.
XXX we could track whether tty_fd has already been successfully opened,
XXX the ttystate initialised, and then just never close it unless it is
XXX necessary, then we can keep COLUMNS/LINES accurate in scripts, even
is not found, from a suggestion by RT (LP: #912691)
• try harder (in a loop) to acquire a file lock if the locking mechanism
documents EINTR is a possibility (fcntl always, flock on Linux not .Ox)
• use -std=c99 not -std=gnu99 if it must be at all
if we find any, but not later; do not check on every read
⇒ allows changing COLUMNS and LINES (independent of each other, or both)
for script shells by passing them in an environment setting, even if
we get a tty; interactive shells still check before each line is read…
reported by the PLD guys, thanks
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
• 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)
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)…
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
• 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
– 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
│Don't alias 'stop' to 'kill -STOP'
│
│Android has already has a stop command used
│to stop the main runtime and the alias
│interferes with testing tools that expect
│stop to kill the runtime.
│
│Change-Id: I02b7efb9203dc39e97f63eb702a54ff79935b316
Although, this is closer to his first patchset and only takes
care of the alias, not the testsuite (which doesn’t run, at
least not out-of-the-box, nicely anyway) using #ifdef ANDROID.
We certainly want a more flexible testsuite…
and vendor pdksh versions, re-introduce FPOSIX alongside FSH. The semantics
are now:
‣ set -o posix ⇒
• disable brace expansion and FSH when triggered
• use Debian Policy 10.4 compliant non-XSI “echo” builtin
• do not keep file descriptors > 2 to ksh
‣ set -o sh ⇒
• set automatically #ifdef MKSH_BINSHREDUCED
• disable brace expansion and FPOSIX when triggered
• use Debian Policy 10.4 compliant non-XSI “echo” builtin
• do not keep file descriptors > 2 to ksh
• trigger MKSH_MIDNIGHTBSD01ASH_COMPAT mode if compiled in
• make “set -- $(getopt ab:c "$@")” construct work
Note that the set/getopt one used to behave POSIXly only with FSH or
FPOSIX (depending on the mksh version) set and Bourne-ish with it not
set, so this changes default mksh behaviour to POSIX!
• 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
• 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
call it only if $RANDOM is indeed set (although pool extension would be a
possibility we do have arc4random_atexit which does it nicely too)
• avoid calling setspec for int→str conversion just before execve()
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
• we must not set the item pointer to NULL, since subsequent ktscan()
would stop there and not find any later occurrences
possible resolution strategies:
‣ still keep tablep; store a dummy value (either (void *)-1 or, probably
more portable, &ktenter or something like that) as is-free marker
⇒ retains benefit of keeping count of actually used entries
⇒ see below for further discussion
‣ don't keep tablep; revert back to setting entry->flag = 0
⇒ need to ktwalk() or ktsort() for getting number of entries
⇒ most simple code
‣ same but with a twist: make ktscan() set pp to the first one with
!(entry->flag & DEFINED)¹ so that it can subsequently be re-used,
or, more accurate, free’d and the entry pointer re-used
⇒ less chance of texpand()ing when not needed
‣ similar (from kabelaffe@): in ktsearch(), move the one we DID find
to the first unused one
⇒ doesn’t need tablep or something, but has the overall best
memory use
⇒ more complicated ktscan(): needs to check pointer for NULL, for
dummyval, then entry->flag
⇒ makes lookup more expensive
⇒ benefit: self-optimising hash tables
⇒ loss: still need ktwalk() or ktsort()
• when afree()ing in ktremove(), …
① need to take FINUSE into account
• Python-2.5.4/Objects/dictnotes.txt talks about cache lines
‣ linear backward scan is much worse than linear forward scan
(even if we have to calculate the upper C-array bound)
‣ dereferencing the entry pointer in ktscan() is a penalty
• Python-2.5.4/Objects/dictobject.c has a lot of comments and
a rather interesting collision resolution algorithm, which
seems to de-cluster better than linear search at not much
more cost
• clib and libobjfw have unusable (for looking-at-for-ideas)
hash table implementations
this is a no-op change breaking ifdef-out-d code; the most likely
to happen is to switch to the following scheme:
• keep tablep in struct tbl
• use a magic pointer value for ktremove’d entries, deallocate
the struct tbl as soon as possible – if not FINUSE, immediately
inside ktremove()
‣ memory gain, despite needing to have tablep around
• nuke ktdelete, so that all ops go through kt{enter,remove}
‣ gains us accurate fill information
‣ speed gain: ktscan() needs no longer dereference removed entries
‣ memory (ktsort) and speed (ktwalk) gain: removed entries are now
ignored right from the beginning, so tstate->left and the size
of the sorted array are accurate
‣ removed entries no longer can cause texpand() to be invoked
⇒ this does not give us self-optimising tables, but a speed and
memory benefit plus, probably, simplicity of code; we accurately
know how many non-deleted entries are in a keytab so we can cal-
culate if we need to expand, how much space ktsort() is going to
need, and, for when indexed arrays will be converted to use key-
tabs instead of singly linked linear lists, ${#foo[*]} is fast
(although ${!foo[*]}² and ${foo[*]}³ will need some tweaking and
may run a little less quickly)
• shuffle code around, so that things like search/scan and garbage
collection can be re-used
• use Python’s collision resolution algorithm ipv linear search
② the list of keys needs to be sorted, at least for indexed arrays⁴
③ this needs to be sorted by keys, at least for indexed arrays⁴
④ … but this is a nice-to-have for associative arrays⁵ as well
⑤ which we however do not have
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
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)
• 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
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
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.
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
tg