tangled.org
commits
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: Anirudh Oppiliappan <anirudh@tangled.sh>
as a result, this lets us avoid making another call to knots.
Signed-off-by: oppiliappan <me@oppi.li>
uses the newly added description filter that only autlinks, and renders
backticks, bold and italics.
Signed-off-by: oppiliappan <me@oppi.li>
this sanitizer only renders a tiny subset of markdown, to be used in
inline elements like PR titles and repo descriptions etc.
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
sqlite db still has `at_uri` column to easily use as reference
Signed-off-by: Seongmin Lee <boltlessengineer@proton.me>
Signed-off-by: Seongmin Lee <boltlessengineer@proton.me>
There is no reason for `ResolvedRepo` to not directly inherit `Repo`
Signed-off-by: Seongmin Lee <boltlessengineer@proton.me>
tailwind seems to apply this to every html element. when applying
font-mono to an element, it inherits the ss01 font feature. this results
in a different `a` characters from when inside `pre` and `code` blocks,
where it resets the list of font-features.
i can't seem to find a good fix for this issue. removing the ss01
feature altogether seems like a good fix, the number alternates in inter
aren't all that different in my opinion.
Signed-off-by: oppiliappan <me@oppi.li>
- remove redundant divs in base layout
- allow scrolling changed-files listing
- show more of the file in tree views
- unify repo-index and repo-tree views
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: Anirudh Oppiliappan <anirudh@tangled.sh>
Signed-off-by: Anirudh Oppiliappan <anirudh@tangled.sh>
Replace `Issue.IssueAt` to `Issue.Rkey`. `issue_at` column still exist,
but not used and will be removed in future.
Remove `db.SetIssueAt()` in flavor of `Rkey`.
Create rkey *before* adding to DB, and save same rkey to PDS.
This is safe as `tx` is will be dropped when `ctx` is canceled with
error.
Signed-off-by: Seongmin Lee <boltlessengineer@proton.me>
Otherwise, when programs create files in $HOME, others that check the Git
repo's state may be sad (or at least emit warnings in response to) it
being dirty.
Signed-off-by: Winter <winter@winter.cafe>
IMO, we shouldn't make errors opt-in to actually show upon push, at least
not if/until we have a UI to show them to the user.
Signed-off-by: Winter <winter@winter.cafe>
This moves away from the old method of creating a container with some
shared volumes to one that most users would expect: any changes made
in one step will be accessible by the following steps, and not only if
they're in the workspace or `/etc/nix`. This also paves the way for a
more generic Docker image engine, as users can do things like `apt install`
without the results being blown away across steps.
Signed-off-by: Winter <winter@winter.cafe>
It's possible for SetupWorkflow to provision resources and also fail, so
this makes sense for robustness.
Signed-off-by: Winter <winter@winter.cafe>
Signed-off-by: Winter <winter@winter.cafe>
Signed-off-by: Winter <winter@winter.cafe>
and fix empty-repo rendering.
Signed-off-by: oppiliappan <me@oppi.li>
From https://datatracker.ietf.org/doc/html/rfc9207#section-2.4 :
"Clients MUST [...] compare the result to the issuer identifier of the
authorization server where the authorization request was sent to. [...]
If the value does not match the expected issuer identifier, clients MUST
reject the authorization response and MUST NOT proceed with the
authorization grant."
Signed-off-by: Mike P <fentiger@mythik.co.uk>
Signed-off-by: Anirudh Oppiliappan <anirudh@tangled.sh>
Signed-off-by: Anirudh Oppiliappan <anirudh@tangled.sh>
This reverts commits 1931b62baa4452616203ccbb356da72688bc4587 and 86d0b4060de074759eff40666b90be86c244ab23.
Signed-off-by: Winter <winter@winter.cafe>
Causes too many regressions, unfortunately:
- https://discord.com/channels/1361963801993285692/1361991850327674932/1406398439830323292
- https://discord.com/channels/1361963801993285692/1361991850327674932/1406401591627747519
Ideally, our templates wouldn't be whitespace-sensitive (example #1),
but let's stop the bleeding while we figure it out.
Signed-off-by: Winter <winter@winter.cafe>
I guess it's not idempotent in some cases? Looks like it's happy with
these, though, as rerunning it after this commit produces no changes.
Hopefully this is a one time thing...
Signed-off-by: Winter <winter@winter.cafe>
Since Zed can't easily use our wrapped Prettier (see previous commit),
and we're now using three formatters, let's just add one command to
format all of our code correctly.
Signed-off-by: Winter <winter@winter.cafe>
This also removes the Zed Prettier configuration, because AFAICT, it
never actually worked, and will not work with how we install Prettier
now. :(
Signed-off-by: Winter <winter@winter.cafe>
This does result in the removal of some options that aren't supported by
the SQLite driver [0], but IMO we can stick with the defaults for these.
[0]: https://github.com/mattn/go-sqlite3#connection-string
Signed-off-by: Winter <winter@winter.cafe>
`sql.DB` pools connections, which means that there's a good chance that
our foreign keys aren't actually being enforced, unless a goroutine
happens to be using the connection that was made when we set `pragma
foreign_keys = 1`.
Signed-off-by: Winter <winter@winter.cafe>
do not return errors from ingesters, this causes the read loop to be
killed.
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
define variants for all families, weights and styles.
Signed-off-by: oppiliappan <me@oppi.li>
based on lowercase IBM Plex Mono `t`, italic variant.
thanks to @t1c.dev!
Signed-off-by: oppiliappan <me@oppi.li>
when attempting to do an authorized request while logged-out, the auth
middleware boots the user to the login page. this page now keeps track
of the 'Referer' in the oauth request. once the oauth callback is
complete, the user is sent back to the page they came from.
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
- `inline-block` on li was too general, this only applies on
`li:has(input)` now
- all styles have been moved to input.css and apply only to children of
prose
Signed-off-by: oppiliappan <me@oppi.li>
- initialize sanitizer once, and reuse for life
- improve policies, and allow sanitizer to hold multiple policies
(this will come in handy, for PR titles, repo description, profiles
description etc.)
- add general safe items to allow list, most of these are generated by
goldmark GFM
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
this simplifies the error collection logic a lot.
Signed-off-by: oppiliappan <me@oppi.li>
Signed-off-by: oppiliappan <me@oppi.li>
Even though nothing in this folder would match the extensions we watch,
probably makes sense to not even waste the cycles doing so.
Signed-off-by: Winter <winter@winter.cafe>
This also switches away from `nixos-shell` in the process as by this
point it wasn't really adding much to our setup except inflexibility.
Signed-off-by: Winter <winter@winter.cafe>
It's unidiomatic, and doesn't allow extension by way of changing the
systemd service options (like is done in the following commit).
Signed-off-by: Winter <winter@winter.cafe>
7a076be1af4677a7246a792fecb463e4e6f1afee, but for the actual file *tree*
template (used anywhere we show a diff of a tree).
Signed-off-by: Winter <winter@winter.cafe>
non-forked repos show a book icon, forked repos show a fork icon
Signed-off-by: Samuel Newman <mozzius@protonmail.com>
adds a fork indicator to the repo lists in the profile, if repo is forked
Signed-off-by: Samuel Newman <mozzius@protonmail.com>
Signed-off-by: oppiliappan <me@oppi.li>
tailwind seems to apply this to every html element. when applying
font-mono to an element, it inherits the ss01 font feature. this results
in a different `a` characters from when inside `pre` and `code` blocks,
where it resets the list of font-features.
i can't seem to find a good fix for this issue. removing the ss01
feature altogether seems like a good fix, the number alternates in inter
aren't all that different in my opinion.
Signed-off-by: oppiliappan <me@oppi.li>
Replace `Issue.IssueAt` to `Issue.Rkey`. `issue_at` column still exist,
but not used and will be removed in future.
Remove `db.SetIssueAt()` in flavor of `Rkey`.
Create rkey *before* adding to DB, and save same rkey to PDS.
This is safe as `tx` is will be dropped when `ctx` is canceled with
error.
Signed-off-by: Seongmin Lee <boltlessengineer@proton.me>
This moves away from the old method of creating a container with some
shared volumes to one that most users would expect: any changes made
in one step will be accessible by the following steps, and not only if
they're in the workspace or `/etc/nix`. This also paves the way for a
more generic Docker image engine, as users can do things like `apt install`
without the results being blown away across steps.
Signed-off-by: Winter <winter@winter.cafe>
and fix empty-repo rendering.
Signed-off-by: oppiliappan <me@oppi.li>
From https://datatracker.ietf.org/doc/html/rfc9207#section-2.4 :
"Clients MUST [...] compare the result to the issuer identifier of the
authorization server where the authorization request was sent to. [...]
If the value does not match the expected issuer identifier, clients MUST
reject the authorization response and MUST NOT proceed with the
authorization grant."
Signed-off-by: Mike P <fentiger@mythik.co.uk>
Causes too many regressions, unfortunately:
- https://discord.com/channels/1361963801993285692/1361991850327674932/1406398439830323292
- https://discord.com/channels/1361963801993285692/1361991850327674932/1406401591627747519
Ideally, our templates wouldn't be whitespace-sensitive (example #1),
but let's stop the bleeding while we figure it out.
Signed-off-by: Winter <winter@winter.cafe>
when attempting to do an authorized request while logged-out, the auth
middleware boots the user to the login page. this page now keeps track
of the 'Referer' in the oauth request. once the oauth callback is
complete, the user is sent back to the page they came from.
Signed-off-by: oppiliappan <me@oppi.li>
- initialize sanitizer once, and reuse for life
- improve policies, and allow sanitizer to hold multiple policies
(this will come in handy, for PR titles, repo description, profiles
description etc.)
- add general safe items to allow list, most of these are generated by
goldmark GFM
Signed-off-by: oppiliappan <me@oppi.li>