Bug 385271
| Summary: | /etc/bash_completion silently crashes, stopping execution of sourced .bashrc and .bash_profile scripts | ||
|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Need Real Name <bugzilla> |
| Component: | bash-completion | Assignee: | Ville Skyttä <ville.skytta> |
| Status: | CLOSED WORKSFORME | QA Contact: | Fedora Extras Quality Assurance <extras-qa> |
| Severity: | high | Docs Contact: | |
| Priority: | low | ||
| Version: | 8 | CC: | axel.thimm, sheltren |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2007-11-19 18:20:14 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
I can't reproduce. When I source my (also based on Fedora defaults) ~/.bash_profile manually, /etc/bash_completion is never even read because the first check in /etc/profile.d/bash_completion.sh notices that it has already been sourced. How have you determined that /etc/bash_completion is sourced when you (re)source ~/.bash_profile and that those two lines are causing problems? Yes... by trial-and-error (using echo statements), I followed the chain as
above. Everything gets sourced down the chain.
Each of the lines noted above caused the crash -- and everything loaded smoothly
when I commented them out. I (obviously) get the same problem when I just source
/etc/bash_completion directly or when I run the lines directly from the bash prompt.
The errors I get are:
bash: BASH_COMPLETION: readonly variable
bash: BASH_COMPLETION_DIR: readonly variable
I don't understand what you mean by "the first check in
/etc/profile.d/bash_completion.sh notices that it has already
been sourced.
The first check is a return if $BASH_VERSION is not set (which it is), so
nothing happens. The second check is whether $BASH_VERSION > 2.04, which it is
so it proceeds to the third check which checks whether /etc/bash_completion is
readable, which it is. So, the file /etc/bash_completion is then sourced.
(In reply to comment #2) > Yes... by trial-and-error (using echo statements), I followed the chain as > above. Everything gets sourced down the chain. I used the same method, and it does *not* get sourced for me. Additionally, directly sourcing /etc/bash_completion multiple times in a row causes no "crashes" either. > I don't understand what you mean by "the first check in > /etc/profile.d/bash_completion.sh notices that it has already > been sourced. > > The first check is a return if $BASH_VERSION is not set (which it is), so > nothing happens. The first check additionally checks if $BASH_COMPLETION is nonempty: # Check for bash (and that we haven't already been sourced). [ -z "$BASH_VERSION" -o -n "$BASH_COMPLETION" ] && return (In reply to comment #3) > directly sourcing /etc/bash_completion multiple times in a row causes no > "crashes" either. However I do see that some debug "echo foo >> /tmp/foo" do stop printing when repeatedly sourcing bash_completion in the section you mentioned. Changes further down bash_completion do take effect despite of this. OK - I see what's wrong. My version of bash_completion.sh doesn't have the second part of the test (-o -n "$BASH_COMPLETION"). When I looked closely, I noticed that yum had pulled in the version of bash-completion from the atrpms repo. I will 'cc' Axel Thimm on this to allow him to fix his version. Axel: you may want to update your version by making sure it has the line: # Check for bash (and that we haven't already been sourced). [ -z "$BASH_VERSION" -o -n "$BASH_COMPLETION" ] && return Sorry for the confusion - I wish there were a way to more explicitly know when one of the other repos overrides a standard fedora rpm :) Thanks for the info. Not sure why ATrpms ships a bash-completion package for Fedora in the first place though. |
Description of problem: When I source my .bash_profile script (based on the Fedora default), it sources the following scripts in sequence: ~/.bash_profile -> ~/.bashrc -> /etc/bashrc -> /etc/profile.d/bash_completion.sh -> /etc/bash_completion. However, the scripts are never fully executed, because the bash_completion script silently crashes on *each* of the following two lines: BASH_COMPLETION="${BASH_COMPLETION:-/etc/bash_completion}" BASH_COMPLETION_DIR="${BASH_COMPLETION_DIR:=/etc/bash_completion.d}" The crashes are silent because stdout and stderr are redirected to /dev/null. The comment says that the purpose is to "silence readonly variable errors". The reality is that the errors abort the script even when silenced. Note this problem doesn't seem to occur when .bashrc is sourced as part of a login or new bash shell process. However, people are used to being able to (re) source their .bash_profile and .bashrc files and this error causes them all to abort!