Bug#893418: Corrupted package names in by_vote.gz
On Fri, Mar 23, 2018 at 02:24:39PM +0100, Enrico Zini wrote:
> On Fri, Mar 23, 2018 at 02:10:07PM +0100, Bill Allombert wrote:
>
> > > Probably. Is the format of that file documented somewhere?
> > This is a list of key/value pair in RFC822 style.
> > See /usr/share/doc/popularity-contest/examples/bin/README.examples
> > for the format of the Package line.
>
> I have a few questions:
>
> How is the package name separated from the integer fields? It does not
> look like a fixed-width field:
>
> Package: abev-form-obhgepi-fpk-nav 0 0 0 2
> Package: abev-form-obhgepi-fpk-nav-egyeb 0 0 0 2
>
> If it is instead space-separated, currently I didn't see package names
> that contained spaces, but is there a guarantee that the package name
> won't contain spaces?
It is garanteed that package name will not contain spaces.
> Alternatively, should the parsing instead be done by splitting on \s+
> from the right with a maximum of 4 splits?
>
> Some package names seem to be truncated, like this one:
>
> Package: apache-openoffice-4.1.4-linux-x86-install-rpm-de 0 0 0 1
The server should not truncate anything. I will check what happened.
> Is the character set guaranteed to be UTF8
Definitely no.
> or should I parse it as
> binary, and drop all lines that do not decode as UTF8, or even all lines
> that are not strictly 7-bit ascii, like this one?
There is no reason to assume that UTF8 is "better" than non-UTF8 here.
Cheers,
--
Bill. <ballombe@debian.org>
Imagine a large red swirl here.
Reply to: