[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Fedora-livecd-list] [patch 00/20] Fix error handling and misc. cleanups



Hey Mark,

Sorry for the lag - been busy trying to get a new HAL release out,

On Tue, 2007-03-27 at 15:14 +0100, Mark McLoughlin wrote:
> Hi,
> 	Apologies for spamming the list with 20 mails, but hopefully
> it'll make it easier to review the patches.

Yeah, it did make it easier although I'm always slighty scared when
seeing a bunch of patches ;-) - would you happen to have a git repo
handy that I can pull from?

> 	The background to this is that I was suprised to see that
> livecd-creator had rather poor error handling (e.g. not unmounting
> stuff after certain errors, obscure error messages etc.) since
> I thought that was the main reason to re-write pilgrim in python :-)
> 
> 	It's understandable, though. Because Python has exceptions,
> it's easy to be fooled into thinking you have error conditions
> under control, but in reality, exceptions probably makes it easier
> to screw up error handling.
> 
> 	So, here's an attempt to fix it based on the following
> guidelines:
> 
>   - don't intercept an exception if you can't do something useful
>     with it. Corollary: intercept an exception if you can do something
>     useful with it.
> 
>   - anticipate user errors or user-fixable system errors in a user
>     friendly manner. Corollary: allow other types of exceptions
>     to be passed to the toplevel.
> 
>   - i.e. there are 3 types of errors/exceptions, and 3 different
>     strategies for handling them:
> 
>       1) errors which we can handle ourselves without ever telling
>          the user - e.g. a directory which we wish to create already
>          exists
> 
>       2) user errors, or predictable system errors, which we
>          should do our best to help the user out with -  e.g.
>          a busted kickstart file or running out of loop devices
> 
>       3) pathological system errors or programmer errors, all
>          of which can be just allowed to propogate up to the
>          toplevel where an ugly traceback will be printed -
>          e.g creating a dir under the temporary build dir
>          fails or an unmount fails because we didn't close
>          the rpmdb
> 
>   - whatever the error, we should do our best to clean up files,
>     directories, mounts, etc. which we created
> 
>   - bear in mind that e.g. Ctrl-C can happen at any time
> 
> 	Also, I've cleanup up various parts of the code as I
> worked on the error handling. Those patches come first.

This is really good stuff, thanks a bunch for working on it. I just read
through all the patches and from a cursory look they look good to me;
tomorrow I'm going to try and actually apply them to the code [1].

I'd also like for Jeremy to sign them off as he was the last one before
you to fix up my, uh, rough bash->python conversion of the pilgrim shell
script plus he's the main guy working on this stuff these days.

Thanks again, Cheers,
David

[1] : As a practical thing would you happen to have a git repo stashed
somewhere that I can pull form?



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]