Declaring an empty global var as a reference in a function twice crashes bash
Configuration Information:
Machine: x86_64
OS: darwin13.3.0
Compiler: gcc
Compilation CFLAGS: -DPROGRAM='bash' -DCONF_HOSTTYPE='x86_64'
-DCONF_OSTYPE='darwin13.3.0' -DCONF_MACHTYPE='x86_64-apple-darwin13.3.0'
-DCONF_VENDOR='apple' -DLOCALEDIR='/usr/share/locale' -DPACKAGE='bash' -DSHELL
-DHAVE_CONFIG_H -DMACOSX -I. -I. -I./include -I./lib -I./lib/int\
l -I/Users/apple/Downloads/bash-4.3/lib/intl -arch i386 -arch x86_64 -O3
-fomit-frame-pointer -momit-leaf-frame-pointer -ffast-math -fvisibility=hidden
-march=core2 -mtune=corei3 -mmacosx-version-min=10.8 -isysroot
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platfo\
rm/Developer/SDKs/MacOSX10.8.sdk
uname output: Darwin Z 13.4.0 Darwin Kernel Version 13.4.0: Sun Aug 17
19:50:11 PDT 2014; root:xnu-2422.115.4~1/RELEASE_X86_64 x86_64
Machine Type: x86_64-apple-darwin13.3.0
Bash Version: 4.3
Patch Level: 26
Release Status: release
Description:
When you have already set a variable to empty, running `declare -gn
varname` in a function twice crash bash with a Segment Fault.
`set -xv` told me that bash would stop right when running declare.
There is also some additional info from lldb.
I have reproduced this on my OS X and Ubuntu 14.10 LTS amd64,
and a bash 4.3.30 on the same OS X machine, in another prefix.
The LLDB output was nearly identical.
Repeat-By:
Run this in bash, either as a script with she-bang or copy-paste in
interactive shell:
```Bash
#!/bin/bash
v=''
a(){ declare -gn v; }
a; a
```
--LLDB--
* thread #1: tid = 0x922581, 0x000100074371 bash`declare_internal + 3729,
queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1,
address=0x29)
frame #0: 0x000100074371 bash`declare_internal + 3729
bash`declare_internal + 3729:
-> 0x100074371: orb$0x10, 0x29(%r14)
0x100074376: movq -0x58(%rbp), %rsi
0x10007437a: movl 0x28(%r14), %edx
0x10007437e: testb $0x2, %dl
[DOC] Documentation for `complete' compacts[] is incomplete.
Configuration Information: Machine: i686 OS: cygwin Compiler: gcc Compilation CFLAGS: -DPROGRAM='bash.exe' -DCONF_HOSTTYPE='i686' -DCONF_OSTYPE='cygwin' -DCONF_MACHTYPE='i686-pc-cygwin' -DCONF_VENDOR='pc' -DLOCALEDIR='/usr/local/share/locale' -DPACKAGE='bash' -DSHELL -DHAVE_CONFIG_H -DRECYCLES_PIDS -I. -I. -I./include -I./lib -O1 uname output: CYGWIN_NT-10.0-WOW VAIO-lynn 2.0.4(0.287/5/3) 2015-06-09 12:20 i686 Cygwin Machine Type: i686-pc-cygwin Bash Version: 4.3 Patch Level: 30 Release Status: release Description: Bash has incomplete documentation for builtins in `complete.def`. The [-abcdefgjksuv] options to specify types to complete are not documented. This is making writing completions some kind of black magic. All those correspondence can be found in complete.def. More specifically, compacts[]. /* By the way is anyone going to assign those '\0's so we can do even more things? */ Repeat-By: help compgen, help complete, info bash, man bash. Fix: Write approporiate doc for this. I only need a transcription of compacts[] so I can look it up without Luking it.
RE: Feature Request re: syslog and bashhist
You can add `-DSYSLOG_HISTORY` to your CFLAGS for building.
And for formatting, apply this patch:
--- bashhist.c 2015-08-11 00:09:38.449468800 +0800
+++ bashhist.c 2015-08-11 00:09:42.970623400 +0800
@@ -713,12 +713,12 @@
char trunc[SYSLOG_MAXLEN];
if (strlen(line) < SYSLOG_MAXLEN)
-syslog (SYSLOG_FACILITY|SYSLOG_LEVEL, "HISTORY: PID=%d UID=%d %s",
getpid(), current_user.uid, line);
+syslog (SYSLOG_FACILITY|SYSLOG_LEVEL, "bash[%d]: UID=%d %s", getpid(),
current_user.uid, line);
else
{
strncpy (trunc, line, SYSLOG_MAXLEN);
trunc[SYSLOG_MAXLEN - 1] = '\0';
- syslog (SYSLOG_FACILITY|SYSLOG_LEVEL, "HISTORY (TRUNCATED): PID=%d
UID=%d %s", getpid(), current_user.uid, trunc);
+ syslog (SYSLOG_FACILITY|SYSLOG_LEVEL, "bash[%d] (TRUNCATED): UID=%d %s",
getpid(), current_user.uid, trunc);
}
}
#endif
I'm not quite familiar with those configure.ac magic, and therefore I can't
write some AC_ARG_ENABLE for you.
Re:Re: [DOC] Documentation for `complete' compacts[] is incomplete.
At 2015-08-11 01:29:42, "Chet Ramey" wrote: >Have you considered reading the man page or perhaps the info doc? That's >where the actions are documented: > > -A action > The action may be one of the following to generate a > list of possible completions: > alias Alias names. May also be specified as -a. > arrayvar > Array variable names. > >...and so on. Oops, I missed it.<>
Enhancement: bash should have uniform escape syntaxes for `echo -e`, `printf` and `$'ANSI_C_style_escape'`.
Configuration Information: Machine Type: i686-pc-cygwin, i686-pc-msys, x86_86-linux-gnu, x86_64-Apple-Darwin Bash Version: 4.3 Patch Level: 30 Release Status: release Description: Bash has different escape syntaxes for `echo -e`, `printf` and `$'ANSI_C_style_escape'`. Take a specific point, `printf` and `$'C_Style'` accepts octals not starting with 0, but `echo -e` doesn't. This is causing quite a lot of confusion. On older versions of bash, like bash 3.2 on OS X, `echo` doesn't understand `\e` but `printf` and `$'C_style'` does. I believe this is quite a aged problem. Repeat-By: printf 'printf starts the bold\33[1m\n'; echo $'C escape should end it\33[0m'; echo -e 'echo -e does nothing but giving crap\33[1;31m'
