On 3/22/10 6:21 PM, Mike Frysinger wrote:
> ok, the trouble is that we take the code, save it via `set`, and then restore
> it.
OK, that was the missing piece. Not really anything to do with the
redirection code per se.
> the assumption is that what bash outputs is equivalent to the original
> code. bash-4 though does not provide equivalent code.
Yeah, that's pretty weird. An artifact of the changes to print here
documents correctly in the presence of operators like |, ||, and &&.
> so here is the reduced test case:
> $ cat test.sh
> foo() {
> rm -f a b c
> for f in a b c; do
> cat <<-EOF >> ${f}
> file
> EOF
> done
> grep . a b c
> }
>
> $ . ./test.sh
> $ foo
> a:file
> b:file
> c:file
The attached patch fixes it for me. Let me know.
Chet
--
``The lyf so short, the craft so long to lerne.'' - Chaucer
``Ars longa, vita brevis'' - Hippocrates
Chet Ramey, ITS, CWRU c...@case.edu http://cnswww.cns.cwru.edu/~chet/
*** ../bash-4.1-patched/print_cmd.c 2009-09-16 15:32:26.000000000 -0400
--- print_cmd.c 2010-03-22 21:15:30.000000000 -0400
***************
*** 114,117 ****
--- 114,123 ----
#define CHECK_XTRACE_FP xtrace_fp = (xtrace_fp ? xtrace_fp : stderr)
+ #define PRINT_DEFERRED_HEREDOCS(x) \
+ do { \
+ if (deferred_heredocs) \
+ print_deferred_heredocs (x); \
+ } while (0)
+
/* Non-zero means the stuff being printed is inside of a function def. */
static int inside_function_def;
***************
*** 561,571 ****
{
print_for_command_head (for_command);
-
cprintf (";");
newline ("do\n");
indentation += indentation_amount;
make_command_string_internal (for_command->action);
semicolon ();
indentation -= indentation_amount;
newline ("done");
}
--- 566,578 ----
{
print_for_command_head (for_command);
cprintf (";");
newline ("do\n");
+
indentation += indentation_amount;
make_command_string_internal (for_command->action);
+ PRINT_DEFERRED_HEREDOCS ("");
semicolon ();
indentation -= indentation_amount;
+
newline ("done");
}
