[arch-dev-public] newest bash and bash-completion
Hey guys, Newest bash seems to kill bash-completion. I custom-wrote a patch and phrakture, toofishes, and I have all had a run at testing it (definitely not extensively). Basically the way it handles setting of read-only vars makes it blow up, and causes the remainder of our .bashrcs (after sourcing bash_completion) to not be executed. I've pushed a patched version of bash-completion to TESTING - if those of you using bash_completion could all make sure it's solid I'd appreciate it. :) Thanks! -- Travis
Travis Willard schrieb:
Newest bash seems to kill bash-completion. I custom-wrote a patch and phrakture, toofishes, and I have all had a run at testing it (definitely not extensively). Basically the way it handles setting of read-only vars makes it blow up, and causes the remainder of our .bashrcs (after sourcing bash_completion) to not be executed.
I've pushed a patched version of bash-completion to TESTING - if those of you using bash_completion could all make sure it's solid I'd appreciate it. :)
What exactly does it kill? I just ran a quick test and it seemed fine. As bash-completion is not that critical, you could have pushed it to extra directly IMO.
On Tue, 18 Sep 2007 09:26:43 +0200 Thomas Bächler <thomas@archlinux.org> wrote:
Travis Willard schrieb:
Newest bash seems to kill bash-completion. I custom-wrote a patch and phrakture, toofishes, and I have all had a run at testing it (definitely not extensively). Basically the way it handles setting of read-only vars makes it blow up, and causes the remainder of our .bashrcs (after sourcing bash_completion) to not be executed.
I've pushed a patched version of bash-completion to TESTING - if those of you using bash_completion could all make sure it's solid I'd appreciate it. :)
What exactly does it kill? I just ran a quick test and it seemed fine.
For both Dan and myself, when the old /etc/bash_completion script was sourced from within our .bashrc files, it would get to the point where it tried to set BASH_COMPLETION and BASH_COMPLETION_DIR (which by then were already readonly variables), then it would just die, outputting to stderr (redirected to /dev/null) a message saying that they were readonly. Then, nothing in our .bashrc files following the source of /etc/bash_completion would get executed.
As bash-completion is not that critical, you could have pushed it to extra directly IMO.
Testing was suggested by both Aaron and Dan. *shrug* -- Travis
Travis Willard schrieb:
For both Dan and myself, when the old /etc/bash_completion script was sourced from within our .bashrc files, it would get to the point where it tried to set BASH_COMPLETION and BASH_COMPLETION_DIR (which by then were already readonly variables), then it would just die, outputting to stderr (redirected to /dev/null) a message saying that they were readonly. Then, nothing in our .bashrc files following the source of /etc/bash_completion would get executed.
I didn't realize any of this, but it sound reasonable. What surprises me is that I only added the latest bugfixes to bash, this shouldn't lead to such breakage.
As bash-completion is not that critical, you could have pushed it to extra directly IMO.
Testing was suggested by both Aaron and Dan. *shrug*
For the exact same reason as above, I didn't push bash to testing. If it works for you two, we can add bash-completion with your fixes to extra.
participants (2)
-
Thomas Bächler
-
Travis Willard