Re: `verb=(--) declare -ax verb` causes an ENOTSOCK segfault
On Fri, Jun 16, 2023, at 1:02 AM, Wiley Young wrote: > Description: >`verb=(--) declare -ax verb` causes bash to crash and closes the active > terminal tab. The value "(--)" does not seem to be necessary: % /bin/bash -c 'foo= declare -ax foo' zsh: segmentation fault /bin/bash -c 'foo= declare -ax foo' I am seeing this behavior with bash 3.2.57, so it appears to be longstanding. -- vq
Re: `verb=(--) declare -ax verb` causes an ENOTSOCK segfault
On Fri, Jun 16, 2023 at 10:59:24AM -0400, Lawrence Velázquez wrote: > % /bin/bash -c 'foo= declare -ax foo' > zsh: segmentation fault /bin/bash -c 'foo= declare -ax foo' > > I am seeing this behavior with bash 3.2.57, so it appears to be > longstanding. I get a segfault/core dump in every version of bash starting from 2.05b, which is the oldest I've got. In case this helps, here's the backtrace from the core from bash 5.2: (gdb) bt #0 0x5582fea45845 in array_insert (a=0x558300ae61b0, i=i@entry=0, v=v@entry=0x558300ae6070 "") at array.c:574 #1 0x5582fea12d3a in bind_variable_internal (name=0x558300ae60f0 "foo", value=0x558300ae6130 "", table=table@entry=0x558300ac3cd0, hflags=hflags@entry=0, aflags=aflags@entry=544) at variables.c:3231 #2 0x5582fea13d2a in push_temp_var (data=0x558300ae7110) at variables.c:4625 #3 0x5582fea3d093 in hash_flush (table=table@entry=0x558300ae7290, free_data=free_data@entry=0x5582fea13e90 ) at hashlib.c:395 #4 0x5582fea178e8 in dispose_temporary_env ( pushf=pushf@entry=0x5582fea13e90 ) at variables.c:4690 #5 0x5582fea1797a in dispose_used_env_vars () at variables.c:4710 #6 0x5582fea0afb9 in execute_command_internal (command=0x558300ae37d0, asynchronous=asynchronous@entry=0, pipe_in=pipe_in@entry=-1, pipe_out=pipe_out@entry=-1, fds_to_close=fds_to_close@entry=0x558300ae3a10) at execute_cmd.c:872 #7 0x5582fea65c19 in parse_and_execute (string=, from_file=from_file@entry=0x5582feab50a7 "-c", flags=flags@entry=20) at evalstring.c:536 #8 0x5582fe9f22a0 in run_one_command ( command=0x7ffd19e26aa0 "foo= declare -ax foo") at shell.c:1473 #9 0x5582fe9f0aab in main (argc=3, argv=0x7ffd19e26628, env=0x7ffd19e26648) at shell.c:763
Re: 5.2 fails to treat a missing conditional expression as an error of syntax
On 6/15/23 8:57 PM, Kerin Millar wrote: As below. $ bash -c 'declare -p BASH_VERSION' declare -- BASH_VERSION="5.2.15(1)-release" $ bash -c '[[ ]]; echo fin'; echo $? 0 https://lists.gnu.org/archive/html/bug-bash/2022-10/msg00103.html describes the fix. We had this same discussion back in February: https://lists.gnu.org/archive/html/bug-bash/2023-02/msg00156.html where I answered in more detail: https://lists.gnu.org/archive/html/bug-bash/2023-02/msg00163.html You participated. -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ``Ars longa, vita brevis'' - Hippocrates Chet Ramey, UTech, CWRUc...@case.eduhttp://tiswww.cwru.edu/~chet/
Re: `jobs` shows output even when nothing has been started in the background
On 6/14/23 6:12 PM, Ajeet D'Souza wrote: Hey Chet, Thanks for the quick reply. I have some follow-up questions: * Is there an elegant way I can start an external process without it showing up under jobs, even on affected versions? My current best idea is: (_="$(/bin/echo)"; jobs -l) You can try to disown it. * Is there an elegant way I can get the correct number of jobs, even on affected versions? My current best idea is to run jobsonce before doing the count: (jobs &> /dev/null; jobs | wc -l) You have to get the jobs you're not interested in removed from the jobs table. That's as good a way as any if you don't know the job number or if it's not the current job. -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ``Ars longa, vita brevis'' - Hippocrates Chet Ramey, UTech, CWRUc...@case.eduhttp://tiswww.cwru.edu/~chet/
Re: `jobs` shows output even when nothing has been started in the background
Hi, Thanks for all the help! I created a PR with this workaround on the Starship repository: https://github.com/starship/starship/pull/5253 Regards, Ajeet On Sat, Jun 17, 2023 at 12:00 AM Chet Ramey wrote: > On 6/14/23 6:12 PM, Ajeet D'Souza wrote: > > Hey Chet, > > > > Thanks for the quick reply. I have some follow-up questions: > > > > * Is there an elegant way I can start an external process without it > > showing up under jobs, even on affected versions? My current best > idea > > is: (_="$(/bin/echo)"; jobs -l) > > You can try to disown it. > > > * Is there an elegant way I can get the correct number of jobs, even on > > affected versions? My current best idea is to run jobsonce before > doing > > the count: (jobs &> /dev/null; jobs | wc -l) > > You have to get the jobs you're not interested in removed from the jobs > table. That's as good a way as any if you don't know the job number or > if it's not the current job. > > > -- > ``The lyf so short, the craft so long to lerne.'' - Chaucer > ``Ars longa, vita brevis'' - Hippocrates > Chet Ramey, UTech, CWRUc...@case.eduhttp://tiswww.cwru.edu/~chet/ > >
bug-bash@gnu.org
On 6/10/23 7:31 PM, Grisha Levit wrote: The command printing code can fail to add a required semicolon when the last word in the command ends with `&' Thanks for the report and patch. Chet -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ``Ars longa, vita brevis'' - Hippocrates Chet Ramey, UTech, CWRUc...@case.eduhttp://tiswww.cwru.edu/~chet/
bug-bash@gnu.org
On 6/10/23 10:41 PM, Martin D Kealey wrote: On Sun, 11 Jun 2023, 09:31 Grisha Levit, wrote: The command printing code can fail to add a required semicolon when the last word in the command ends with `&' This could be obviated by unconditionally outputting a newline instead of a semicolon. There's some merit to this; the existing group command printing code already prints group commands on multiple lines if within a shell function. I acknowledge that this style isn't to everyone's taste, but to me it feels more "natural" to the Shell, rather than pretending that one is writing in some other language; I detest style guides that insist on obscuring ";then" on the end of a long line. Everyone has a preference. -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ``Ars longa, vita brevis'' - Hippocrates Chet Ramey, UTech, CWRUc...@case.eduhttp://tiswww.cwru.edu/~chet/
Re: posix_spawn (or vfork) support?
On 6/15/23 8:32 PM, Sam James wrote: Hi, Sorry if this has come up before - I did take a look and couldn't find anything. Could bash use posix_spawn/vfork instead of the rather heavyweight fork? I haven't looked at them. Bash does more work between fork and exec than the posix_spawn implementations I've looked at can manage, so it's not a real win. I've never looked at vfork because there are a lot of assumptions baked into the code that the child has its own copies of variables and other data. -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ``Ars longa, vita brevis'' - Hippocrates Chet Ramey, UTech, CWRUc...@case.eduhttp://tiswww.cwru.edu/~chet/
Re: 5.2 fails to treat a missing conditional expression as an error of syntax
On Fri, 16 Jun 2023 14:25:54 -0400 Chet Ramey wrote: > On 6/15/23 8:57 PM, Kerin Millar wrote: > > As below. > > > > $ bash -c 'declare -p BASH_VERSION' > > declare -- BASH_VERSION="5.2.15(1)-release" > > $ bash -c '[[ ]]; echo fin'; echo $? > > 0 > > https://lists.gnu.org/archive/html/bug-bash/2022-10/msg00103.html > describes the fix.1 > > We had this same discussion back in February: > > https://lists.gnu.org/archive/html/bug-bash/2023-02/msg00156.html Thanks. > > where I answered in more detail: > > https://lists.gnu.org/archive/html/bug-bash/2023-02/msg00163.html > > You participated. Good heavens. This entirely slipped my mind. -- Kerin Millar
[PATCH] fix compgen -V leak
My earlier patch for adding compgen -V did the variable assignment in a
pretty silly way and had a small memory leak to boot. Hope this new way
makes sense, sorry for the extra work.
---
>From b6b13b89e1436ddd575483a81e79ef43d82a5c0c Mon Sep 17 00:00:00 2001
From: Grisha Levit
Date: Sat, 3 Jun 2023 16:37:02 -0400
Subject: [PATCH] fixup compgen -V array handling
* builtins/complete.def
- compgen_builtin: skip the unnecessary WORD_LIST conversion, avoid
small leak
* externs.h
- remove strlist_to_word_list which was only added for initial version
of compgen -V support
---
builtins/complete.def | 7 +--
externs.h | 1 -
2 files changed, 1 insertion(+), 7 deletions(-)
diff --git a/builtins/complete.def b/builtins/complete.def
index 890cf20d..bb03c6e9 100644
--- a/builtins/complete.def
+++ b/builtins/complete.def
@@ -678,7 +678,6 @@ compgen_builtin (WORD_LIST *list)
int old_ind, old_completion, old_quoting, old_suppress;
SHELL_VAR *var;
char *varname;
- WORD_LIST *alist;
if (list == 0)
return (EXECUTION_SUCCESS);
@@ -763,11 +762,7 @@ compgen_builtin (WORD_LIST *list)
var = builtin_find_indexed_array (varname, 1);
if (var && sl && sl->list && sl->list_len)
{
- alist = strlist_to_word_list (sl, 0, 0);
- assign_array_var_from_word_list (var, alist, 0);
- free (sl);
- sl = (STRINGLIST *)NULL;
- dispose_words (alist);
+ array_from_argv (array_cell(var), sl->list, sl->list_len);
rval = EXECUTION_SUCCESS;
}
}
diff --git a/externs.h b/externs.h
index a1363d4d..fe5aa492 100644
--- a/externs.h
+++ b/externs.h
@@ -414,7 +414,6 @@ extern STRINGLIST *strlist_prefix_suffix (STRINGLIST *,
const char *, const char
extern void strlist_print (STRINGLIST *, const char *);
extern void strlist_walk (STRINGLIST *, sh_strlist_map_func_t *);
extern void strlist_sort (STRINGLIST *);
-extern WORD_LIST *strlist_to_word_list (STRINGLIST *, int, int);
/* declarations for functions defined in lib/sh/stringvec.c */
--
2.41.0
[PATCH] uninitialized variable access
Some uninitialized variable access identified by clang's static analyzer.
(FWIW 90% of the reports were bogus but these seem legit)
* lib/readline/input.c
- rl_gather_tyi: the `result' variable is no longer initialized before
first access since commit d0bc56a32
* lib/readline/kill.c
- _rl_read_bracketed_paste_prefix: if there's no further input after the
initial \e of the bracketed paste prefix, an uninitialized value of `key'
can get used in the return value
* subst.c
- function_substitute: seems like the `tflag' assignment ended up in the
wrong place?
* builtins/read.def
- read_builtin: with `read -n0', the initialization of saw_escape is goto-d
over but that variable is later accessed
---
diff --git a/lib/readline/input.c b/lib/readline/input.c
index 229474ff..00605834 100644
--- a/lib/readline/input.c
+++ b/lib/readline/input.c
@@ -252,6 +252,7 @@ rl_gather_tyi (void)
chars_avail = 0;
input = 0;
tty = fileno (rl_instream);
+ result = -1;
/* Move this up here to give it first shot, but it can't set chars_avail
*/
/* XXX - need rl_chars_available_hook? */
diff --git a/lib/readline/kill.c b/lib/readline/kill.c
index 1dfe3c57..1f13e447 100644
--- a/lib/readline/kill.c
+++ b/lib/readline/kill.c
@@ -779,7 +779,7 @@ _rl_read_bracketed_paste_prefix (int c)
pbpref = BRACK_PASTE_PREF; /* XXX - debugging */
if (c != pbpref[0])
return (0);
- pbuf[ind = 0] = c;
+ pbuf[ind = 0] = key = c;
while (ind < BRACK_PASTE_SLEN-1 &&
(RL_ISSTATE (RL_STATE_INPUTPENDING|RL_STATE_MACROINPUT) == 0) &&
_rl_pushed_input_available () == 0 &&
diff --git a/subst.c b/subst.c
index 08d9285e..e69e0e5b 100644
--- a/subst.c
+++ b/subst.c
@@ -7021,7 +7021,6 @@ function_substitute (char *string, int quoted, int
flags)
/* We call anonclose as part of the outer nofork unwind-protects */
BLOCK_SIGNAL (SIGINT, set, oset);
lseek (afd, 0, SEEK_SET);
- tflag = 0;
istring = read_comsub (afd, quoted, flags, &tflag);
UNBLOCK_SIGNAL (oset);
}
@@ -7029,6 +7028,7 @@ function_substitute (char *string, int quoted, int
flags)
{
s = get_string_value ("REPLY");
istring = s ? comsub_quote_string (s, quoted, flags) : savestring
("");
+ tflag = 0;
}
run_unwind_frame ("nofork comsub"); /* restores stdout, job control
stuff */
diff --git a/builtins/read.def b/builtins/read.def
index cb4e1e59..80d1241d 100644
--- a/builtins/read.def
+++ b/builtins/read.def
@@ -403,6 +403,9 @@ read_builtin (WORD_LIST *list)
input_string = (char *)xmalloc (size = 112); /* XXX was 128 */
input_string[0] = '\0';
+ pass_next = 0; /* Non-zero signifies last char was backslash. */
+ saw_escape = 0; /* Non-zero signifies that we saw an escape char */
+
/* More input and options validation */
if (nflag == 1 && nchars == 0)
{
@@ -463,9 +466,6 @@ read_builtin (WORD_LIST *list)
add_unwind_protect (xfree, rlbuf);
#endif
- pass_next = 0; /* Non-zero signifies last char was backslash. */
- saw_escape = 0; /* Non-zero signifies that we saw an escape char */
-
if (tmsec > 0 || tmusec > 0)
{
/* Turn off the timeout if stdin is a regular file (e.g. from
[PATCH] null pointer deref in bindpwd
Only triggered by doing something stupid:
bash -c 'declare -n OLDPWD=X[SHLVL=-1]; /; cd /'
bash: line 1: X[SHLVL=-1]: bad array subscript
Segmentation fault: 11
---
diff --git a/builtins/cd.def b/builtins/cd.def
index de123f8b..e3156463 100644
--- a/builtins/cd.def
+++ b/builtins/cd.def
@@ -158,10 +158,9 @@ bindpwd (int no_symlinks)
pwdvar = get_string_value ("PWD");
tvar = bind_variable ("OLDPWD", pwdvar, 0);
- if (tvar && readonly_p (tvar))
+ if (tvar == 0 || readonly_p (tvar))
r = EXECUTION_FAILURE;
-
- if (old_anm == 0 && array_needs_making && exported_p (tvar))
+ else if (old_anm == 0 && array_needs_making && exported_p (tvar))
{
update_export_env_inplace ("OLDPWD=", 7, pwdvar);
array_needs_making = 0;
