From: Matt Jolly
Hi all,
Thanks to dev-libs/openssl exposing an RFC 9000 (QUIC) API from 3.2,
cURL is able to use OpenSSL as a QUIC backend in addition
to its current suppport via net-libs/ngtcp2.
Considering the current state of QUIC implementations, the popularity
of cURL, its support for a m
From: Matt Jolly
The CURL_QUIC USE_EXPAND enables us to sanely manage QUIC (RFC 9000)
backends as they are added to cURL in the future: currently there are
two supported implementations, OpenSSL and ngtcp2, however it's likely
that other popular TLS libraries will expose QUIC APIs over time,
and
From: Matt Jolly
Signed-off-by: Matt Jolly
---
profiles/base/make.defaults | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/profiles/base/make.defaults b/profiles/base/make.defaults
index c70c5d198531..957af187bda2 100644
--- a/profiles/base/make.defaults
+++ b/profiles/base/
From: Matt Jolly
Thanks to dev-libs/openssl exposing a QUIC API from
3.2, cURL is able to use OpenSSL as a QUIC backend.
This commit enables users to select between the
`curl_quic_ngtcp2` and `curl_quic_openssl` backends
via their associated USE_EXPAND and provides a mechanism
to easily add more
From: Matt Jolly
We're the ones deviating from the norm. Including the profiler does not
noticably increase the build time (at least on amd64), nor does it
pull in additional dependencies.
What it _does_ do is break upstream build systems that assume
a standard rust configuration like upstream w
From: Matt Jolly
Inherit the rust eclass and take advantage of eclass features like
`RUST_MIN_VER`.
Also replace calls to `cargo` with the rust eclass exported ${CARGO}.
If used without llvm-r1 (typical usage) an || dependency on Rust
slots (between `RUST_M{AX,IN}_VER` if set) will be added.
I
From: Matt Jolly
This patchset slots Rust, adds `rust.eclass`, a helper eclass for slotted Rust,
and updates `cargo.eclass` to take advantage of the new slotted Rust, requiring
a
revbump of dependent packages.
The intent is to immediately deprecate and shortly remove legacy Rust slots
and the v
From: Matt Jolly
This requires a version of LLVM that is not keyworded for mips.
Signed-off-by: Matt Jolly
---
profiles/arch/mips/package.use.mask | 4
1 file changed, 4 insertions(+)
diff --git a/profiles/arch/mips/package.use.mask
b/profiles/arch/mips/package.use.mask
index fa77510aa1
From: Matt Jolly
Use the llvm-r1 eclass and our knowledge of the internal
LLVM used in -bin to add an appropriate `llvm_slot_{x}`
IUSE to each Rust-bin ebuild so that the Rust (and LLVM)
eclasses can select an appropriate slotted implementation.
`LLVM_OPTIONAL` is used as we don't need to consum
From: Matt Jolly
Signed-off-by: Matt Jolly
---
.../chromium/chromium-130.0.6723.91-r1.ebuild | 1387 +
1 file changed, 1387 insertions(+)
create mode 100644 www-client/chromium/chromium-130.0.6723.91-r1.ebuild
diff --git a/www-client/chromium/chromium-130.0.6723.91-r1.ebuild
From: Matt Jolly
Firefox is a bit unusual in that the LTO via the USE=clang path
requires a strong dependency between the selected Rust and LLVM
slots, while building with GCC (and without LTO) does not.
Leverage llvm-r1 to add `llvm_slot_x` USE and use these to ensure
that an appropriate Rust i
From: Matt Jolly
The rust eclass acts similarly to the llvm eclass.
It works with optional `RUST_{MAX,MIN}_SLOT` variables to
enable ebuilds to trivially generate (and enforce) dependencies
on an appropriate Rust SLOT.
A `RUST_NEEDS_LLVM` variable can be set to have the eclass read
`LLVM_COMPAT
From: Matt Jolly
Signed-off-by: Matt Jolly
---
.../rustls-ffi/rustls-ffi-0.13.0-r1.ebuild| 107 ++
1 file changed, 107 insertions(+)
create mode 100644 net-libs/rustls-ffi/rustls-ffi-0.13.0-r1.ebuild
diff --git a/net-libs/rustls-ffi/rustls-ffi-0.13.0-r1.ebuild
b/net-libs/
From: Matt Jolly
Signed-off-by: Matt Jolly
---
gnome-base/librsvg/librsvg-2.58.2-r1.ebuild | 421
1 file changed, 421 insertions(+)
create mode 100644 gnome-base/librsvg/librsvg-2.58.2-r1.ebuild
diff --git a/gnome-base/librsvg/librsvg-2.58.2-r1.ebuild
b/gnome-base/librsv
From: Matt Jolly
Updating vulnerable (or otherwise outdated) crates in Rust ebuilds
is painful. Generally speaking, there are 5 options:
- Run `cargo update` to fetch new versions from the web.
This is obviously not suitable for use in Portage.
- Patch the software via Portage to accept a non-
From: Matt Jolly
Updating vulnerable or outdated crates in Rust ebuilds is not easy.
This patchset aims to reduce the developer effort and knowledge
required to substitute crates into a Rust build process.
We do this by:
- Enabling `paths = ['/foo/bar/baz']` style global replacements.
This us
From: Matt Jolly
These are the the old ebuilds, updated to be slotted.
Other changes:
- `USE=rls` was dropped at some point, this is now
hard-enabled.
- As the LLVM is far too old for Gentoo we are using the bundled LLVM
exclusively.
In particular, dev-lang/rust:1.54.0 forms an important p
From: Matt Jolly
Signed-off-by: Matt Jolly
---
app-antivirus/clamav/Manifest | 2 +-
.../{clamav-1.0.7-r1.ebuild => clamav-1.0.7-r2.ebuild}| 8 +++-
2 files changed, 8 insertions(+), 2 deletions(-)
rename app-antivirus/clamav/{clamav-1.0.7-r1.ebuild => clama
From: Matt Jolly
These variables enable users (though most likely Gentoo developers)
to override the selection of the Rust implementation by the eclass.
This means that _only_ the specified ERUST_SLOT and/or ERUST_TYPE
('source' or 'binary') will be checked for, with the eclass `die`ing
with "No
From: Matt Jolly
This commit addresses some feedback that the Rust eclass should provide
feedback to users on what it's actually doing and why a particular
Rust was (or was not) deemed suitable for use.
To do this we now:
- Explicitly note if a Rust slot was skipped due to LLVM_SLOT
incompati
From: Matt Jolly
The `mips` logic has been moved into the one ebuild that it impacts,
and that ebuild is only required for one package - we'll try to
remove it from the tree soon.
Merge `loong` into `rust_all_arch_uris()` output unconditionally;
the check is now superfluous - all pre-1.71.0 rust
From: Matt Jolly
Also update `rust_arch_uri` to also fetch the asc if we're renaming
so that we can use verify-sig.
Signed-off-by: Matt Jolly
---
eclass/rust-toolchain.eclass | 34 +++---
1 file changed, 19 insertions(+), 15 deletions(-)
diff --git a/eclass/rust-to
From: Matt Jolly
These commits are part of a larger body of work which will
enable us to add the `beta` and `nightly` channels to
dev-lang/rust{,-bin}.
Please feel free to provide feedback on the commits directly
via the following PR: https://github.com/gentoo/gentoo/pull/40524
which will be mer
23 matches
Mail list logo
