While I was there, I also switched gofmt's vendor ignoring logic to use
git's builtin pathspec filtering.
Change-Id: Iab8093d795e7f967cceef351c6a593ba776d9790
to db3ee9ee8931e51af7158114b9d02bf93d1a7811
and also update golang.org/x/crytpo to b47b1587369238182299fe4dad77d05b8b461e06
The golang.org/x/crypto change is to correct a typo in 'Gopkg.toml'
which resulted in the 'Gopkg.lock' not containing an actually locked
down revision.
The leveldb change fixes#1185
Change-Id: If14a0c4d6fc011716c2371351f2572b0ca2b588a
In order to facilitate manual recovery in severe scenarios, the
--keep-going flag will make the server come back online even when there
are index or recovery errors. This will allow further querying on the
unhealthy server to identify missing data or to extract data for
migration to a new/healthy instance.
Fixes#1166
Change-Id: I4153905757d9e7d8014780dd5660a862cb8ba1ab
To support composition, all optional interfaces in Go need a way to
say, at runtime, that "oh, nevermind, I defensively implemented this
interface but it turns out the thing I'm wrapping doesn't support it".
We learned that lesson only in the past few years, but SubFetcher
predates that.
So do that here and make proxycache implemented SubFetcher and make
everything respect the fallback oh-nevermind error value.
Fixes#1170
Change-Id: I3c0958cf6692a324e98f526571d10a352a8e18b4
I have tested that, starting with an empty $GOPATH/bin, after running
devcam test, $GOPATH/bin still has none of the Perkeep related binaries.
Fixes#1139
Change-Id: I60b63ff917cf91c8d424b448f874f40a18e1c23b
The old directory aspect (directory_detail.js), was done before the
current (React-based) look and layout of the web UI. It is still
functional, but does not look very good, and does not match the look of
the rest of the web UI.
Therefore, this change adds a new directory layout, with a
blob container approach and layout similar to the one for permanode
sets, and replaces the old one as the default Directory aspect.
It seems to have performance issues (slow loads), but I think it's
better to enable it now, rather than wait again for many months to
land it. Especially if it forces us to address said performance issues.
If it's too bad and users want the old one in the meantime, it should be
easy to re-enable the old as the default.
Change-Id: Ieb0cde8290f0a058008cdc73b596ad1a330ded1a
My previous commit (f3f38f0c76) made a change that https became true
if httpsCert or httpsKey was enabled. This broke devcam server, which
always sets the cert+key fields in config/dev-server-config.json but
always set "https" to either true or false.
Instead, make it an error to omit "https" if the cert or key is
specified. (It's still currently legal to specify https==false but
specify a cert or key files, so devcam server now works, but it's not
clear that that's a good idea. I might revisit this later when I
update dev-server-config.json to the upcoming TOML format.)
Fixes#1168
Change-Id: I29d000c5eb9472cc901fbd4cb91159a6c293570e
This will allow consumers of pkg/schema to do things like verify the
signature on claims.
Fixes#1167
Change-Id: Iec76e382e18d91f82694412bc855ac5be82df825
They were named conn because nearly 8 years ago the pre-ResponseWriter
type was named *http.Conn:
https://golang.org/doc/devel/weekly.html#2010-09-29
Time to modernize.
Change-Id: I68d1a5f16dfa6fc2e7a1f863e5d9edb7308527dc
Niklas Merz