This bug is largely so I don't forget about it, but a couple of things occured
to me about fsck which should really be addressed at some stage:
1. The question asking code shouldn't require a <return> after asking a y/n
question so that its compatible with other fsck question asking interfaces.
2. Is there any reason we cannot rename the binary to fsck.gfs2 like all the
other filesystems? Obviously we need backwards compatibility, but a symlink in
the packaging should deal with that.
1. Should be easy.
2. Should be easy.
I'd like to add #3:
When I made the changes to allow SIGINT to be caught and to ask the user
if he/she wants to skip the current pass, I got into a bind whereby
if you interrupt it while a question is being asked, it does not
re-ask the question, so the user may be left confused. My solution at the
time was to ignore the interrupt while it's asking any Y/N questions.
The problem with that is obvious: If there is a massive amount of
fs corruption, it will keep asking annoying questions and never let
the user break out of it. Unless, of course, they go to another
terminal cli and use the kill command. That's bad because sometimes
they may just want to get out and restart it with the -y option.
Shouldn't be too difficult, but not quite as easy as #1 and #2.
So #3: Allow <ctrl-c> interrupts to be processed in a friendly
manner while a Y/N question is being asked.
Created attachment 160826 [details]
Proposed patch to fix the problem
This patch has been tested on system trin-10 by manually answering
questions with a single character (no enter), interrupting the various
passes, and interrupting the questions.
I posted the proposed fix to cluster-devel on 08 August 2007.
Nobody commented on the proposed patch so I'm assuming it's good to
go for RHEL5.2.
Tested on system trin-10. Committed to cvs in HEAD and RHEL5 branches.
Changing status to Modified.
splitting out bug 286211: gfs2_fsck not found by fsck wrapper
Discovered a small problem with this fix. Need a modify it slightly.
Resetting to assigned state, since it hasn't been built yet and still
doesn't have the right ack flags at this time.
Created attachment 224531 [details]
This addendum patch fixes the problem mentioned in comment #6.
Created attachment 224551 [details]
Addendum patch revised
Fixed a stupid bug in the previous addendum patch.
Fix tested on roth-01 then committed to HEAD and RHEL5 branches for
inclusion in 5.2. Setting back to modified state.
An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on the solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.