e141cab1ac
revert cid 10050DD0C8519194AE9 and 100511FC81525949AF7
...
AT&T SVR4 broke 'cc -c -o foo.o foo.c' totally
switch to an alternative method of finding out the prototypes
provided in IRC by dalias (thanks); RT says this fixes his
issues on SVR4 – let’s just ignore the ominous message from
cid 1004713D70A5362BACF about needing to be a compile-time
check unless something else pops up, I can’t think of something
that isn’t already covered
2013-02-17 07:06:15 +00:00
b684a7c6bd
backpedal with $'…' and $"…" interpolation
...
it turns out this breaks more legacy scripts than anticipated
2013-02-17 05:40:17 +00:00
47f8a21e3c
SVR4 cc doesn’t like -c -o conftest.o when conftest.o already exists:
...
it errors out with “-o would overwrite conftest.o” (thanks RT for reporting)
2013-02-16 17:55:51 +00:00
85d7059c48
… I fsck’d up and built R42b from MAIN ipv mksh-R42stable… oh well.
...
TODO: convert enum to something like uint8_t to save even more space
2013-02-16 00:21:57 +00:00
6b797a7bc8
• test.sh now is -v (verbose) by default, new option -Q to quieten (like Build.sh)
...
• test.sh handles +P, +Q, +S, +v properly
2013-02-11 16:27:56 +00:00
42e1bdc095
doc update
2013-02-10 21:20:57 +00:00
43a41932e7
make DEBUG_LEAKS safely free all resources before the main shell exits
2013-02-10 18:17:30 +00:00
b23d768a9b
Unbreak linking on AIX, reported (with fix) by Torsten Sillke, 10x!
2013-01-12 02:25:01 +00:00
72a748c891
if you have to look *this* up, danger ahead; point to porter's info
2013-01-01 21:19:36 +00:00
0ce1ca41cb
carefully begin owcc (Watcom C’s POSIX cc wrapper) support
2012-12-28 04:58:13 +00:00
113db497ba
RT says QNX 4 has _SIGMAX not SIGMAX or NSIG
2012-12-28 03:35:34 +00:00
192ea7638b
now that we found out the root case for the recent QNX problem…
...
03:11⎜«RT|AO:#!/bin/mksh» YES it WORKS now!
… there’s no reason for sys_nerr to not be const, either ☺
2012-12-28 03:20:35 +00:00
f7acd1f3fc
AIEEE!
...
03:00⎜«RT|AO:#!/bin/mksh» cc -c -o a.out does not output a.out but conftest.o (!!)
evil… just use tcfn=conftest.o for now, cf. cid 1004713D70A5362BACF (but why must it be one?)
2012-12-28 03:05:17 +00:00
b2396677e7
harmonise sys_{sig,err}list ⓐ checks, ⓑ uses, ⓒ _decl values when not needed, ⓓ prototypes; ⓔ “const” is a keyword and thus space-separated from the preceding ‘*’ pointer indicator
2012-12-28 02:28:39 +00:00
64eddba65f
-D__EXT_UNIX_MISC needed for “extern const char * const sys_errlist[];” on QNX
2012-12-27 15:52:47 +00:00
f998054508
16:39⎜«RT|Chatzilla:#!/bin/mksh» ok, tested cc -E outputs preprocessed out with leading space(!).
...
16:49⎜<mirabilos:#!/bin/mksh> that's like NeXTstep which insers spaces around =
16:49⎜<mirabilos:#!/bin/mksh> but we can sed that away I guess.
16:49⎜«RT|Chatzilla:#!/bin/mksh» yeah, modifying grep/sed lines adding " *" works
2012-12-24 17:50:10 +00:00
6d5e27a31a
gcc-snapshot (see below) issued a clobber warning, and both it and mgcc
...
yowled about the memmove test until I found a compromise
gcc version 4.8.0 20121120 (experimental) [trunk revision 193662] (Debian 20121120-1)
2012-12-22 00:03:42 +00:00
8e6fa9a5c6
now we can actually prefer sys_errlist[] (oh, and _sys_errlist[] also exists, a̲n̲d̲ fix a pasto) over strerror(3) like we do for sys_siglist[] and strsignal(3); our implementation is smaller and does more code reuse than even MirBSD libc’s after all
2012-12-17 23:31:30 +00:00
12c662b6e2
thanks to ISC (Interactive) Unix, we now know a few samples of errnos
...
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
2012-12-17 23:18:11 +00:00
ae1ec32dbc
fix whitespace; ACK RT’s Minix-vmd stmts
2012-12-17 22:57:50 +00:00
4c5d7094ef
RT also said what was missing on SunOS 4.1.1 (it also needs -DMKSH_UNEMPLOYED?)
2012-12-17 22:14:27 +00:00
b8d3f12025
merge/genericise RT’s port to Minix-vmd
2012-12-17 21:55:06 +00:00
75af4e244a
add -fno-asynchronous-unwind-tables for all GCC builds, as suggested by dalias and chris2
2012-12-05 19:38:05 +00:00
7fc9e1b492
revert cid 100506499941EC578A6: musl has _ALL_SOURCE now, aliased to
...
_GNU_SOURCE which we already define for convenience; thanks to dalias
and chris2 on IRC for discussion and being accomodating!
2012-12-03 22:10:02 +00:00
2213c5887a
how many times can you botch a gettimeofday mirtoconf check?
2012-12-03 13:07:11 +00:00
6e41c9c2ee
catch MSYS dash printf builtin not working
2012-12-01 14:22:09 +00:00
b868c517b7
remove *all* nonnull assertions and other workarounds for clang scan-build
...
this beast evolved in the last 2½ years, and we’ve had trouble with
some of them earlier
2012-12-01 01:36:30 +00:00
f6728353bd
fix “null argument where non-null required” warning
2012-11-30 22:17:32 +00:00
ce6cdb96f2
regenerate Makefiles
2012-11-30 20:49:18 +00:00
0f3071a8b2
MKSH_DISABLE_EXPERIMENTAL is a NOP again; use ${ precmd;} in dot.mkshrc
2012-11-30 20:19:16 +00:00
2fc5aa487b
RT tells me Minix 2 (i386) also doesn’t have gettimeofday(2)
2012-11-20 18:50:46 +00:00
bfe7d78d40
bring back ${ foo;} sans dot.mkshrc patch, using a temporary file, and as experimental feature
2012-10-22 20:19:18 +00:00
574c024635
introduce MKSH_DISABLE_EXPERIMENTAL and wrap the new feature introduced
...
in cid 1005084678C510CF7E4 in it
2012-10-22 16:53:22 +00:00
59f7279ea2
this is now http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55009
2012-10-21 17:38:22 +00:00
21234f1526
fix detection of penguin flocks
2012-10-14 14:02:10 +00:00
d8662eb228
even more hacks to pass “-O666 -fstrict-overflow -Wstrict-overflow=9 -flto=jobserver” with “gcc version 4.8.0 20120930 (experimental) [trunk revision 191865] (Debian 20120930-1)” ☹ plus make the rtchecks mandatory
2012-10-03 17:24:23 +00:00
0d72035990
make char_is_8_bits assertion more anal
2012-09-28 18:57:49 +00:00
4b2371eebf
add workaround for musl (define _BSD_SOURCE as well) which is even *more*
...
ugly than the one for dietlibc (the other one misunderstanding the *real*
meaning of that flag), by $CC divining, until such time as all the Linux
C libraries will honour _ALL_SOURCE (I wish)…
2012-09-27 18:23:43 +00:00
89613d34d4
detect zsh 2.5.02 as found on NeXTstep 3.3; thanks to ft (Frank Terbeck) and chris2 (Christian Neukirchen)
2012-09-18 14:18:59 +00:00
84020897fd
allow overriding /etc location (LP#1039713), but don’t do it
2012-08-24 19:09:10 +00:00
1d21782419
add -D_DARWIN_C_SOURCE on OSX; should help bsiegert@’s problem
2012-08-03 18:16:43 +00:00
4778a2a62f
lesson learned from http://k1024.org/~iusty/blog/entry/perf-null/ – add -DMKSH_SMALL_BUT_FAST which gives more speed (8/20K less cycles, 5/9K less insns, 1.8/2.4k less branches, 65/275 less branch misses) on Debian/amd64 (klibc-static/eglibc) at cost of 0/2 more page faults and 6K/6K more text size
2012-07-01 15:55:00 +00:00
978d365ab1
-L implies convfds now
2012-07-01 15:51:24 +00:00
c1f821d4e5
(mksh) tighten 32-bit requirements; (lksh) switch to long; allow any bitness
2012-06-28 20:17:39 +00:00
1fa4453946
mh… all I’m gonna hack on upstream today… fix -t for manpage generation and cleanup code snippets; bump vsn; sync clog
2012-06-26 19:33:33 +00:00
967b68e08e
one more thing to implicitly disable (nonsensical anyway) for lksh
2012-06-26 19:15:11 +00:00
967e929e98
add basic LEGACY KSH mode (Build.sh -L), no changes yet
2012-06-25 16:05:10 +00:00
b0543ef064
define $tfn default late (for Things To Come™)
2012-06-24 20:47:07 +00:00
3ebfd6e563
new option -t (target shell filename)
2012-06-24 20:46:07 +00:00
4c65f518af
RT says: libc5 uses __inline__ for its inline stuff, but lcc doesn't support __inline__ but __inline
2012-05-18 01:38:04 +00:00
1eebfa8b85
check whether klibc sigsuspend() is usable: RT or hpa’s two new methods; if not, deactivate it
2012-05-17 19:36:41 +00:00
6b2aa91ddd
add a way to detect varying C libraries or other stuff “extra type”
...
(for now klibc) alongside “compiler type”
if klibc, add -DMKSH_NO_LIMITS automatically
2012-05-17 19:14:07 +00:00
d05c7c4867
recognise (but don’t do anything with, yet) lcc (from RT)
2012-05-17 18:54:28 +00:00
b89b96a4ab
RT says not using sigsuspend helps Syllable (but is still buggy, see http://forum.syllable.org/viewtopic.php?p=8171 and probably (doesn’t work for me) http://pyro-os.org/?section=Forum&forum=Bugs&zpbw_postKey=NUgSMp85XGHKbF751zJ6Ww which show the bug affects all shells)
2012-05-09 23:20:42 +00:00
1117a25748
• sync clog
...
• write more notes on klibc, BeOS, Coherent, PW32, Syllable, Xenix
• sort DOS OE section to where it belongs
• add TARGET_OS=_svr4 for Dell UNIX
2012-05-07 17:03:03 +00:00
d9d2485905
overhaul the grep(1)s and sed(1)s – see re_format(7)
2012-05-06 15:40:31 +00:00
ea8f29f428
sprinkle CONSERVATIVE_FDS for many older OSes; DISABLE_TTY_WARNING for BeOS and Coherent (probably more to come); NO_CMDLINE_EDITING to Plan 9; put Hurd NO_PATH_MAX into MKSH__ private namespace
2012-05-05 17:37:44 +00:00
27387673a4
new MKSH_DISABLE_TTY_WARNING for ports that just can’t help it
...
TODO: add_cppflags this automatically in the TARGET_OS switch for some
2012-05-04 22:34:51 +00:00
e6ace01f12
last parts of Coherent patchkit: O_ACCMODE and termio
2012-05-04 22:18:27 +00:00
fef3808126
more symlink(7) nonexistence support code
2012-05-04 22:05:02 +00:00
401116766b
generalise the skipping sete{u,g}id() case for Coherent, not just BeOS
2012-05-04 21:57:38 +00:00
5488e79d3b
introduce a nosymlink check category (idea by RT); pass it to test.sh; use cp if ln -s fails (me)
2012-05-04 21:42:51 +00:00
b1aa05e36e
do not use $(uname -s || uname) as TARGET_OS if it’s the same as $(uname -n), to weave around many ancient unicēs; idea discussed with RT
2012-05-04 21:37:08 +00:00
000d0eb653
another awk(1) compatibility fix from RT
2012-05-04 21:34:12 +00:00
bc61602efc
RT points out there may be a tab inside the NSIG expansion
2012-05-04 21:33:14 +00:00
cb7dc2b45b
fix copy/paste accident; found by RT
2012-05-04 21:30:06 +00:00
2d0ca10629
patch test.sh to support check.t split into multiple chunks
...
which is apparently needed on Coherent due to OS limits
inspired by a patched test.sh from RT, except I require a
dot before the sequence number for easier globbing, made
the splitting use -S, and the errorlevel is more correct
2012-05-04 21:28:06 +00:00
5780bc5eb7
new MKSH_NO_CMDLINE_EDITING to disable command line editing in its entirety
...
mainly for the Plan 9 port though it may also help the WinAPI variant,
other porting efforts, as well as a new project I cannot say yet
2012-05-04 20:49:08 +00:00
b67de6ba17
add experimental code to use sigprocmask+pause+sigprocmask ipv sigsuspend (and harden j_sigchld handler for that) to improve working on BeOS 5.0 and Coherent UNIX, found by RT
2012-04-27 16:16:23 +00:00
48da83288d
work around segfault bug in GNU sed 2.03, spotted by RT on Debian 0.91
2012-04-16 17:49:40 +00:00
ea6b0c60f5
write a bit about pcc, 386BSD, klibc
2012-04-14 19:35:43 +00:00
5eea59ad57
rewrite maketemp() obsoleting tempnam(3) and mkstemp(3) external deps
2012-04-14 16:07:48 +00:00
b497da1eb4
add 386BSD, thanks to the unbelievable RT
2012-04-14 14:11:07 +00:00
b49abfac18
genericise MKSH_NO_SIGSETJMP
2012-04-14 14:07:47 +00:00
0e0b379615
sometimes <stdlib.h> doesn’t provide NULL but <stddef.h> does (386BSD)
2012-04-14 14:04:13 +00:00
29e4269f33
if the OS does not provide sig_atomic_t and/or ssize_t, you can now define
...
MKSH_TYPEDEF_SIG_ATOMIC_T and MKSH_TYPEDEF_SSIZE_T via CPPFLAGS to the
*correct* definitions (it’s absolutely critical they be correct!)
2012-04-14 14:02:40 +00:00
536bc0463f
make -fwrapv common; at least clang also can do it
2012-04-08 20:02:33 +00:00
5b4d0dbbc5
drop all deprecated code, you have been warned
2012-04-07 11:19:53 +00:00
0af5e1f288
can’t forget LynxOS, although it’s still preliminary…
2012-04-06 23:10:50 +00:00
3d3081115b
if LDSTATIC is not empty (i.e. "-static" or " "), always add “our” sources:
...
• strlcpy
• utf_wcwidth
note strchr/strstr from misc.c are still #ifdef DEBUG only, as they are
not eligible: they’re for const-cleanliness debugging purposes
XXX get rid of multiple occurrences of binary search code, too…
2012-04-06 15:20:42 +00:00
ab1ba32d0b
bring back $LDSTATIC support
2012-04-06 15:03:42 +00:00
30949fe090
use a reverse test for sys_siglist_decl – this one works on Dell UNIX
2012-04-06 13:25:51 +00:00
4a89ae1cd0
drop the need for strcasestr: just uppercase a copy and compare with uppercased
2012-04-06 12:59:28 +00:00
5e12953cfe
bring back -DMKSH_UNEMPLOYED default on BeOS as it’s less usable otherwise
2012-04-06 12:57:53 +00:00
6ca4228e7c
disabling jobs just when there is no ctty isn’t needed, at least for BeOS
...
XXX recheck this with all the others, e.g. NeXTstep
2012-04-01 17:48:24 +00:00
3b2d9322ec
change signal probing order: prefer e.g. SEGV over BUS when same Nr. (BeOS)
2012-04-01 17:17:45 +00:00
c30bf73a56
BeOS has no tty nor uid/gid other than 0, a̲n̲d̲ a broken /bin/sh
2012-04-01 16:55:16 +00:00
412a96d29a
speed up Generating list of signal names (XXX can we wrap this into one compile, anyway?)
2012-04-01 16:40:26 +00:00
607a2af260
found the actual reason… cc -E does syntax check and wrap with whitespace ☹
...
also we can only do NeXTstep 3.3 with -posix, but not 3rdpty egcs or OS 4.2
2012-04-01 04:57:24 +00:00
a3fd4f62e6
• move behavioural changes topmost (locale, zsh emulate sh, solaris xpg)
...
• attempt to handle cpp on nextstep by using -save-temps
• work around another nextstep bug that may or may not hit us, found by RT:
‣ http://lists.gnu.org/archive/html/autoconf/2003-11/msg00094.html
• hide "dirname: command not found", scary even if it might help debugging
2012-04-01 02:35:33 +00:00
16732f7e0f
add NeXTstep, OpenStep notes
2012-03-31 18:33:16 +00:00
ac18ec3752
one more case of cpp(1) wanting C syntax on NeXTstep
2012-03-31 17:42:58 +00:00
307224f143
Use “proper C code” for determining the compiler type even though we use
...
only cpp for this: NeXTstep things otherwise:
] conftest.c:42: illegal external declaration, missing ;' after `gcc'
| NeXT Computer, Inc. version cc-437.2.6, gcc version 2.5.8
| NeXT DevKit-based CPP 3.1
| GNU Objective-C version 2.5.8 (80386, BSD syntax) compiled by GNU C version 2.5.8.
2012-03-31 17:37:03 +00:00
4e8c5c078d
RT suggests to drop sys_siglist and _sys_siglist on *all* SCO products, for now, and whitelist those that do need it (no strsignal)
2012-03-30 10:24:45 +00:00
9cac4c92f6
RT wants to make the awk used configurable, apparently OpenStep 4.2 has a broken default one
2012-03-30 09:27:19 +00:00
cf75e7b6ce
couple of minor/cosmetic fixes from RT’s compile farm:
...
• promote SCO OpenServer and UnixWare to !oswarn
• omit trying -O2/-O on OpenServer 5 and USL C
• cast mksh_ari_t to int, mksh_uari_t to unsigned int for printf
• skip ulimit-1 on syllable (which is still too broken)
• write ((mksh_ari_t)-2147483648) ipv UB ((mksh_ari_t)1 << 31)
and add a comment that that is actually meant
• rewrite functions returning !void ending in NOTREACHED
so they’ve got a jump target returning an error at the
end, to aid older compilers and just to be safe
• cast struct stat.st_size to off_t or size_t explicitly when needed
• shorten struct env by two bytes and an alignment, at least
also, optimise control flow and fix more paren matching cases
2012-03-29 19:23:01 +00:00
25745883a9
move USL C detection to after GCC detection, apparently the latter
...
masquerades as the former on SCO systems
2012-03-28 11:15:04 +00:00
df8881b9c8
fix embarrassing typo (and I thought I had tested them all), 10x RT
2012-03-28 11:14:20 +00:00
914c1d816f
drop call to "custom", SCO says to use it to identify system,
...
but RT says it opens a GUI window, and we obviously cannot have this
2012-03-28 11:13:45 +00:00
6dd03ddfad
pass the version to avoid stale Makefile.inc files
2012-03-27 23:13:42 +00:00
tg