Age | Commit message (Collapse) | Author |
|
Having macro identifiers match the "real" identifier style makes macro
code markedly more readable.
Change-Id: Ib675390cc04941ce782c75b2ee686709441081fc
|
|
Change-Id: I025b5f707b15933c4dccdf36ce9e2c8d8673fe74
|
|
Change-Id: Iaf6d5717e2d83a6f81fda34da086311d21f345c2
|
|
This doesn't specifically handle
* the distinction between borrowed/owned file descriptors
(in non-return positions, that only applies to `close`)
* "nullable" fd parameters (-1 for not-present)
* other sentinel fds, like AT_FDCWD
* fd validity in general
Change-Id: I00fa68ec94b8f24e28c661d0c007becb37c0a5f3
|
|
Change-Id: Icfcab2fccc1c152a694dfba3ba1e272cdacdbfab
|
|
Change-Id: I56731f205e065bf501fd31a69c71c10aa7a2f720
|
|
Change-Id: I092d4f47231d560a0c4ac023e63051def34c2c92
|
|
Change-Id: I26356c9b88c838e3c87e4083ac17f958c7c57bb7
|
|
Squeeze the bitflags / SyscallArg impl boilerplate into one macro.
Change-Id: I5b5213b6f3ee11bc61b6f8439ceb2510ef6dffec
|
|
Change-Id: I1b4e7e01b5d5c840ac2e94f9637552f71e61ed04
|
|
Every single syscall arg producing a warning is just noisy.
Change-Id: I339260886110e2982217ebbf9d29cb1cf91c747c
|
|
We separate type-based parsing of syscall arguments from more detailed
validation based on multiple parameters or on specific values.
The former is implemented with a macro that takes syscall signatures
and generates the relevant enum variant and parsing code.
Co-authored-by: edef <edef@unfathomable.blue>
Change-Id: I7e334d3e128e7b1461bfd6fae7c8ec5dd6ada0a8
|
|
Same bug as Ied63822001c8700fc71c89ec16d18036fbc33972, hiding missing
behaviour around dirfd. newfstatat(AT_FDCWD, path, …) is used, as is
newfstatat(fd, "", …).
Change-Id: I03b9dc5118d7a47514abef82c4a577e8362fbb4d
|
|
Our conversion always mangled it, *and* our comparison was inverted,
so both bugs remained hidden.
Change-Id: Ied63822001c8700fc71c89ec16d18036fbc33972
|
|
Change-Id: I74844b6a35fe1fd76af87fe994a346ee92eb3ccb
|
|
Change-Id: I53feb0e3f20c4e9d59407d17612551be4db06177
|
|
This is a bit crude, and it'll behave incorrectly on orphan branches,
but it'll do the trick for now.
Change-Id: Id5424dc4e8480ba65a029ffe2100de0c975e14a1
|
|
We have a single repository (the monorepo), and a single project
channel worth reporting to (Ripple's).
Change-Id: I17d84b610639d7eed422d6a7032908d21143a582
|
|
Change-Id: Ia01dd3e8f8548030d407c3afd1116a5c9eff6ec7
|
|
Change-Id: I41e2932131c31346a421761d45fdc9d2a9b6ff3d
|
|
Change-Id: I066a1fdbeabef0c2c38dcb0740c89eb056bb62f5
|
|
Change-Id: I4670dd7930a9b6227993c1caf92c5877918f982d
|
|
The fourth argument is passed in r10, not rcx, which is merely
clobbered by syscalls.
Change-Id: I0c7210c621f6955deccf21348ce6f495424555cc
|
|
There ought to be a git hook for this..
Change-Id: I111c6150f024782838e40a1855d0d303c6ddbc93
|
|
Change-Id: I31567a95f4d55e7cee526eb5432201087aaa3a1a
|
|
Change-Id: I9e800f347cfb5440c384766f0be55aba4a40c94f
|
|
Change-Id: I9ef820d2c22d54c6d8d6b259c2eea3f27d6459ac
|
|
Change-Id: I107a0402464f13afe318bfb91229dfeacda61806
|
|
Shuffling 3k around on the stack doesn't make anyone happy.
Change-Id: I444fc22267380d9b99ca63ca148b9a9e85238b5a
|
|
This adds a randomly generated Change-Id to every commit message that
doesn't have one yet.
Change-Id: Ic0d1bb74dc60633cd70550e0dde6b0b8340c89bd
|
|
A place for git hooks to live, with an installation script that wires
up core.hooksPath correctly.
Change-Id: Icf6250764cf580067871580d74f6d5ddc52820eb
|
|
With this, minitrace has all the syscalls needed to run cc1.
Change-Id: Iaeef58769f69d64e192af2270eb520e891daec94
|
|
We were accidentally grabbing things from the prelude module already,
and this is the cleaner fix.
Change-Id: I2c6b669164df11f467eae495d109e096eb267b59
|
|
A little more explicit, and a bit more readable.
Change-Id: Id462c46236e1de547aabd36409260bd1c99f6f88
|
|
This is the BuzHash content-defined chunker extracted from Lennart
Poettering's casync, almost verbatim, along with Rust bindings so
we can actually use it safely.
The bindings are a little bit constrained, since they don't support
streaming use, unlike the code they are wrapping. This suffices for
our use case: performance and correctness comparison to our native
Rust implementation, which should behave identically and perform
better.
Change-Id: Iee4296f661b036620bafef5a6c6671fe1a750bb6
|
|
We haven't picked a license for our web content yet, so this is
LicenseRef-NONE for now.
Change-Id: Iabfc35860cb8ad0025624161214b1cbb6574e0c5
|
|
fusermount(1) gets shadowed by pkgs.fuse, but we actually need the
SUID-root one in the system PATH. This adds a wrapper script that
hunts down a SUID-root binary from PATH.
Change-Id: Icc5c6789d7b74bf8f6e3c50529333eb6b894527f
|
|
Unless Caddy recently changed how it resolves directives, this has been
broken since the initial deployment. `error` directives are processed
before `respond` directives[1], and unlike CSS, matcher precision does
not impact directive precedence. Sticking a `route` directive around the
two of them makes things work as expected.
[1]: https://caddyserver.com/docs/caddyfile/directives#directive-order
Change-Id: Ic05c6ba9a7143c3a58c2bf7ccd0498a7337257f8
|
|
cgit-pink purports to be actively maintained, unlike its upstream.
Change-Id: I4d76324ddb24f5fd3e53aeed6fa6ed17713ee15d
|
|
Change-Id: Ie5542d079ee7b2de06e0faed52343905b3287b39
|
|
This will live alongside the existing keys until I'm completely
comfortable with the new setup.
Change-Id: Ia26299c0a1ff60fff37104c0fb8eaf31f17a3df5
|
|
Blocking inline scripts was causing Gerrit to load fonts from Google's
CDN. Rather than adding the appropriate exception for this one instance,
and giving myself a false sense of safety (which will inevitably result
in me running into other subtle issues of this kind in the future), I'm
going to disable the entire thing until I have time to set up reporting.
Change-Id: I7c48e4f7d113ecc15dec0bb930918ccc691b124f
|
|
Change-Id: Ibf68b5b4d7377ea5863315ffd5b6ed24c2874961
|
|
This will host a Gerrit instance and run CI jobs for us.
Change-Id: Ida683bfa910843cc4bcc8a96f2872364067a17ea
|
|
Change-Id: I8329f14bddaaa2d141e82c087821e4602bd1259a
|
|
Change-Id: I6c4d471b2840613d30ee95efe14547f2f9c0db6b
|
|
Change-Id: I0710d80e34bcad1d4b1406731f2d790a0e6972f4
|
|
Node::find would walk the tree all the way down and to the right if
given an out of bounds index. This doesn't affect the return value
(None), but it does waste some cycles walking down all the way to
the deepest rightmost leaf.
Change-Id: Ic2f72aa96291a25819fc6c3c2f060fe0182a7663
|
|
Change-Id: I319a2de0da0ff71f0f337e5a17ef199f23254b11
|
|
A single linear array does everything we need here, since we don't
actually use the cheap writes the BTreeMaps would permit us, and we
save ourselves the hassle of maintaining two parallel lookup structures.
Change-Id: I418b0aaa7a3191fab3195f36f2c68ac0f5f0382b
|