patchwork: mbox download may be mangled version of original mail

[Peter Maydell]

On 7 October 2011 02:57, Jeremy Kerr <jk at ozlabs.org> wrote:
> As you've noticed, the mbox link is not the original mail - patchwork
> doesn't keep that around after parsing the mail, mainly to avoid
> duplicating data.
>
> The mbox is, in effect, a git-am-able version of the plain patch, rather
> than an archive of the original mail. Note that the mail may not be
> am-able in its original form.
>
> However, it would be better to clear up this unnecessary difference. If
> there's a way to do this without duplicating too much data, that would
> be best. Maybe even linking to that entry on a mailing-list archive?

I'd rather patchwork kept the data itself -- one of the attractions
of it to me is that it is a patch archive and means the project
isn't reliant on having a mailing list archive which doesn't
mangle emails.

If you don't want to keep the data twice, you could keep the original
email only, and re-parse and re-generate the web pages/generated-mbox
files on demand. (That would have the added advantage that if bugs
in the parsing code were fixed they'd automatically apply even to
older patches in the archive.)

There's also potential for a DWIM feature where patchwork tries
to fix up line wrapping errors -- I think they should be heuristically
detectable and repairable most of the time :-)

(I ran into this again today for
http://patchwork.ozlabs.org/patch/120638/ -- "lousy mailer" damage
is distressingly common even in otherwise good patches.)

-- PMM