On Fri, May 18, 2012 at 04:17:29PM +0300, Alek Paunov wrote: > Hi Rich, > > On 18.05.2012 15:49, Richard W.M. Jones wrote: > >The reason is that I want to check out all the source for an > >experiment, and I need to know what order of disk space I will need [:-)] > > Would you like to elaborate - What kind of experiment you are preparing? > > [Asking because I live with the idea about Fxx-snapshot mass source > indexing/XRref for long time already (based mainly on the great > David Malcolm's gcc-python-plugin, the openjdk compiler and several > other tools) and if your work will be in the same direction I am > willing to take part of the tasks - e.g. DBs related work ...] It's more of a thought rather than a fully worked out idea, but here you go: Often when something goes wrong, the only thing you have is an error message. Programmers don't always write good errors, and even when they do they are often very context specific (ie. they mean something only in the context of the source code that generates the message, meaning there're a lot of hidden assumptions behind an error). Also it's likely that some errors are encountered by our users very frequently, whereas the vast majority of errors are never printed at all. So [if those assumptions are true] can we do something more useful about error messages? Some ill-formed ideas I had: - Take the text of an error message and display the piece of code which generated it. Not great for end users, but useful for programmers, whom I think of as Fedora's core audience. - Extract error messages from the code and find out how common they are (eg. in Google searches or Bugzilla). This would give us some idea of which error messages should be fixed first to make them more explanatory, or simply bugs which are important to fix. - Some sort of web service for error messages. Google is pretty poor for looking up errors. Try searching for this in Google: "WARNING: gnome-keyring:: couldn't connect to: /run/user/rjones/keyring-IrcfPu/pkcs11: No such file or directory" Currently there are no hits, even though this is a common and long- running bug (RHBZ#783568). The reason is that Google doesn't understand the structure of the error message - the fact that some parts are common to all errors in this class, and others (the pathname) is specific to my machine. But a web service might be able to do better, especially if it has knowledge extracted from the Fedora source code. You could imagine the source code annotated with references to the relevant Bugzilla number. (BTW I'm not claiming exclusivity to any of these ideas. If they sound interesting, please jump in and implement them :-) Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming blog: http://rwmj.wordpress.com Fedora now supports 80 OCaml packages (the OPEN alternative to F#) http://cocan.org/getting_started_with_ocaml_on_red_hat_and_fedora -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel
