bash built-in read timeout (-t) doesn't work

[Miroslav Prašil]

This should timeout after 2 seconds (based on bash man page) but it doesn't:
#read -t 2 test
(waits infinitely, only terminates after pressing [enter] or ctrl-C, etc..)

bash --version:
GNU bash, version 4.2.25(1)-release (x86_64-pc-linux-gnu)

Installed bash package version:
4.2-2ubuntu2.1

$ lsb_release -rd
Description: Ubuntu 12.04.4 LTS
Release: 12.04

[Miroslav Prašil (cezz)]

Not sure if important, but this is running on a virtualbox virt. machine. I've just tried to run the same on EC2 machine with same package version and it seems to be working fine. (timeouts as expected) So this *might* be machine specific, though I don't have any other issues and this seems really odd.

[Miroslav Prašil (cezz)]

I've just noticed I have libreadline5 installed on misbehaving system (due to some dependencies) That's the only difference I can see right now. (correctly working EC2 instance only has libreadline6 installed)

$ dpkg -l *readline*
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name Version Description
+++-================-================-================================================
un libreadline-comm <none> (no description available)
un libreadline-ruby <none> (no description available)
un libreadline4 <none> (no description available)
ii libreadline5 5.2-11 GNU readline and history libraries, run-time lib
ii libreadline6 6.2-8 GNU readline and history libraries, run-time lib
un libterm-readline <none> (no description available)
un libterm-readline <none> (no description available)
ii readline-common 6.2-8 GNU readline and history libraries, common files
un tclreadline <none> (no description available)

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Charles Evans (crtiger)]

Still fails randomly
@ jammy bash 5.1-6ubuntu1
Read -t 60 failure can often be triggered by changing the system clock slightly. It often fails if waiting when ntp runs.

[Lukas Märdian (slyon)]

I cannot seem to reproduce this issue on Jammy bash 5.1-6ubuntu1:
```
$ time read -t 2 test

real 0m2,000s
user 0m0,000s
sys 0m0,000s
$
```

@crtiger Could you please describe your "changing the system clock and wait for ntp run" reproducer in more detail?

[Charles Evans (crtiger)]

While date
Do
Read -t 60 x
Done

In another terminal, adjust the system clock by a few seconds
Repeat hourly until the read hangs.
Sometimes it hangs immediately after setting clock,
Sometimes it runs for hours.
I have seen hangs semi regularly for the last 5, maybe 10+ years.
It often hangs with no apparent system clock changes,
But more often when a cron job runs ntp
No VM involved.
Hangs on multiple AMD systems.

[Launchpad Janitor (janitor)]

[Expired for bash (Ubuntu) because there has been no activity for 60 days.]

[Charles Evans (crtiger)]

Also failing in
Bash 5.0.6ubuntu1.2
In focal 20.04.5 LTS

[Charles Evans (crtiger)]

With multiple bash scripts running for over a month,
I now have random read fails at least once an hour PER SCRIPT.
Restarting the script did not help.
Restarting bash helps for a while.

[Charles Evans (crtiger)]

Hwclock -s is run on the system at least weekly as system time drifts. Adjtime is not installed.

[Charles Evans (crtiger)]

System time consistently falls behind, so hwclock -s is advancing the system clock.