On Thu, 2019-09-26 at 08:53 -0700, Kees Cook wrote:
> On Thu, Sep 26, 2019 at 08:14:03AM -0700, Joe Perches wrote:
> > On Wed, 2019-09-25 at 11:40 -0700, Kees Cook wrote:
> > > Is "6" a safe lower bound here? I thought 12 was the way to go?
> > []
> > > $ git log | egrep 'Fixes: [a-f0-9]{1,40}' | col2 | awk '{print length }'
> > > | sort | uniq -c | sort -n | tail
> > > 238 8
> > > 300 7
> > > 330 14
> > > 344 6
> > > 352 11
> > > 408 40
> > > 425 10
> > > 735 16
> > > 1866 13
> > > 31446 12
> > >
> > > Hmpf, 6 is pretty high up there...
> >
> > Yes, but your grep then col2 isn't right.
> > You are counting all the 'Fixes: commit <foo>' output
> > as 6 because that's the length of 'commit'.
>
> the [a-f0-9]{1,40} already excludes "commit".
No it doesn't as commit starts with c which matches [a-f0-9]{1,40}
Try your original egrep command line yourself.
Maybe use:
$ git log | egrep 'Fixes: [a-f0-9]{1,40}' | awk '{ if (length($2) == 6) { print
$0;} }'
The first few matches are
the commit referenced in Fixes: below replaced the call to
Fixes: commit 18a992787896 ("ARM: ux500: move soc_id driver to drivers/soc")
Fixes: commit 0580dde59438 ("ASoC: simple-card-utils: add
asoc_simple_debug_info()")
Since Fixes: 8c5421c016a4 ("perf pmu: Display pmu name when printing
Fixes: commit 961fb3c206dc ("ASoC: rockchip: rk3399_gru_sound: don't select
unnecessary Platform")
> > I also think the length of the hex commit value doesn't
> > matter much as it's got to be a specific single commit
> > SHA1 anyway, otherwise the commit id lookup will fail.
>
> Fail enough. We do already have 6-digit SHA1 collisions, so it seemed
> like using more than 6 would be nicer? *shrug* I don't have a strong
> opinion. :)
>
> > > > > @@ -1031,6 +1040,7 @@ MAINTAINER field selection options:
> > > > --roles => show roles (status:subsystem, git-signer, list, etc...)
> > > > --rolestats => show roles and statistics (commits/total_commits, %)
> > > > --file-emails => add email addresses found in -f file (default: 0
> > > > (off))
> > > > + --fixes => for patches, add signatures of commits with 'Fixes:
> > > > <commit>' (default: 1 (on))
> > >
> > > Should "Tested-by" and "Co-developed-by" get added to @signature_tags ?
> >
> > All "<foo>-by:" signatures are added.
>
> Ah, I'd missed where that happened. I do note that's only when
> git-all-signature-types is set, which is default 0. (/me goes to add
> this to his invocations...)
>
> my $email_git_all_signature_types = 0;
> ...
> if ($email_git_all_signature_types) {
> $signature_pattern = "(.+?)[Bb][Yy]:";
> } else {
> $signature_pattern = "\(" . join("|", @signature_tags) . "\)";
> }
>
> > > @commit_authors is unused?
> >
> > Yes, authors are already required to sign-off so
> > it's just duplicating already existing signatures.
>
> Sure, it just seemed odd to populate it if it wasn't going to be used.
It's a generic function.
