0001984: No trap handlers are called when wait is interrupted (with possible fix)

BusyBox Bug and Patch Tracking

Viewing Issue Simple Details [ Jump to Notes ]

[ View Advanced ] [ Issue History ] [ Print ]

Category

Severity

Reproducibility

Date Submitted

Last Update

0001984

[BusyBox] Other

major

always

01-28-08 13:47

02-10-08 11:04

Reporter

Ronald Wahl

View Status

public

Assigned To

BusyBox

Priority

normal

Resolution

fixed

Status

closed

Product Version

Summary

0001984: No trap handlers are called when wait is interrupted (with possible fix)

Description

The following shell sccrips does not behave correctly:

----------------------------------
trap_handler () {
    echo trap
}

trap trap_handler USR1

sleep 3600 &
while(true); do wait; done
----------------------------------

Expected behavior: Sending a USR1 signal to the shell should cause "trap" to be printed each time the signal is sent.

Actual behavior:
  bash: works as expected
  1.01: works as expected
  1.1.x ... 1.4.x: only the first signal is honoured
  1.5.x ... 1.9.0: nothing happens

The bug came with SVN revision 18198 of ash.c:
------------------------------
@@ -324,11 +331,13 @@
onsig(int signo)
{
     gotsig[signo - 1] = 1;
- pendingsigs = signo;
+ pendingsig = signo;

     if (exsig || (signo == SIGINT && !trap[SIGINT])) {
- if (!suppressint)
+ if (!suppressint) {
+ pendingsig = 0;
             raise_interrupt();
+ }
         intpending = 1;
     }
}
------------------------------

If I remove the "pendingsig = 0;" before raise_interrupt(); it starts working again.

I'm not sure if this is the right fix. Can anyone comment on this?

Additional Information

Affected versions: busybox > 1.01

The bug in busybox 1.1.x ... 1.4.x is due to a missing sigsetmask(0) which was
removed.

Starting with busybox 1.5.0 the sigsetmask(0) was added again in a more modern form using setprocmask.

Attached Files

Relationships

Notes
(0003534) Ronald Wahl 01-28-08 14:22	Some additional suggestion: Probably swapping pendingsig = 0; raise_interrupt(); is a better approach?

(0003539) vda 01-28-08 15:50	"pendingsig = 0" was added because of this issue: http://busybox.net/bugs/view.php?id=1280 [^]

(0003544) vda 01-28-08 15:55	Swapping "pendingsig = 0;" with "raise_interrupt();" is wrong: raise_interrupt() never returns.

(0003549) Ronald Wahl 01-28-08 16:18 edited on: 01-28-08 16:18	if (signo == SIGINT && !trap[SIGINT]) pendingsig = 0; solves my issue and bug 2 in issue 1280.

(0003824) vda 02-10-08 11:04	Fixed in revision 20973. Thanks.

Issue History
Date Modified	Username	Field	Change
01-28-08 13:47	Ronald Wahl	New Issue
01-28-08 13:47	Ronald Wahl	Status	new => assigned
01-28-08 13:47	Ronald Wahl	Assigned To	=> BusyBox
01-28-08 14:22	Ronald Wahl	Note Added: 0003534
01-28-08 15:50	vda	Note Added: 0003539
01-28-08 15:55	vda	Note Added: 0003544
01-28-08 16:18	Ronald Wahl	Note Added: 0003549
01-28-08 16:18	Ronald Wahl	Note Edited: 0003549
02-10-08 11:04	vda	Status	assigned => closed
02-10-08 11:04	vda	Note Added: 0003824
02-10-08 11:04	vda	Resolution	open => fixed