Fixes <https://bugs.gnu.org/30184>.
Regression introduced in 297e04d660.
Reported by Christopher Baines <mail@cbaines.net>.
* guix/scripts/publish.scm (nar-response-port): Add 'compression'
parameter and honor it.
(http-write): Get 'x-nar-compression' from the initial RESPONSE.
Suggested by atw on #guix.
* gnu/services/shepherd.scm (shepherd-service-file): Use 'scheme-file'
instead of 'gexp->file'.
(shepherd-configuration-file): Likewise, and adjust to non-monadic
style.
(shepherd-boot-gexp): Adjust accordingly.
* guix/scripts/system.scm (upgrade-shepherd-services): Use
'lower-object' in addition to 'shepherd-service-file'.
Fixes <https://bugs.gnu.org/30155>.
The effect can be seen in the package graph produced by:
guix graph -e '(@@ (gnu packages commencement) static-bash-for-glibc)'
This reduces the number of "duplicate" nodes in this graph, i.e.,
distinct package objects that correspond to the same derivation (objects
that are not 'eq?' but semantically equal.)
* gnu/packages/commencement.scm (linux-libre-headers-boot0): Make an
'mlambda' instead of a 'lambda'.
(hurd-core-headers-boot0): Ditto.
Mitigates <https://bugs.gnu.org/29881>.
Reported by Diego Nicola Barbato <dnbarbato@posteo.de>.
* guix/ui.scm (load*): Unset '%fresh-auto-compile' only on Guile 2.2.3.
Previously we were looking at the load of the past 5 minutes, which
means that, after a build, we could end up waiting for 5 minutes for
that metric to be low enough.
* guix/scripts/offload.scm (machine-load): Compute RAW based on ONE, not
FIVE.
Wrap files in bin/ and sbin/ with the location of the gem itself and the
location of any other gems in use (GEM_PATH). This ensures that the bin files
will run with the right environment when executed.
It does however mean that native-inputs will also get wrapped up in any
binaries, which is not good, as it increases the size of the closure, and
risks this code being used at runtime.
* guix/build/ruby-build-system.scm (wrap): New procedure.
(%standard-phases): Add the wrap phase.
A modified copy of wrap-program from (guix build utils). The wrap-program
procedure doesn't work well for Ruby scripts, as it breaks using the -S flag
with ruby to execute the script, as when -S is passed to ruby, it expects the
script on the PATH to use ruby in the shebang, and not bash.
Therefore, to wrap the program, but keep the shebang as ruby, wrap it with a
ruby script instead.
wrap-ruby-program uses .real/foo rather than .foo-real, as this might be
neater. This procedure also includes a call to Gem.clear_paths to make it
possible to set the GEM_PATH through this method, and for it to take effect.
* gnu/build/ruby-build-system.scm (wrap-ruby-program): New procedure.
* guix/build/ruby-build-system.scm (install): Install gems to the vendor
directory, rather than the GEM_HOME. The vendor directory does not include
the version of ruby used to install the gem in the path, which makes it
easier to add it to the GEM_PATH for all versions of ruby to use.
(gem-home): Remove procedure.
* gnu/packages/ruby.scm (ruby, ruby-2.1)[native-search-paths]: Switch to
lib/ruby/vendor_ruby.
(ruby-1.8)[native-search-paths]: Remove native-search-paths.
(gem-directory): Remove procedure.
(ruby-ansi, ruby-ae)[arguments]: Remove use of gem-directory.
(ruby-metaclass, ruby-instantiator, ruby-introspection, ruby-mocha,
ruby-nokogiri, ruby-minitest-tu-shim, ruby-redcloth)[arguments]: Remove use
of gem-home.
(ruby-git, ruby-httpclient)[arguments]: Remove use of GEM_HOME.
* gnu/packages/databases.scm (es-dump-restore)[arguments]: Remove use of
GEM_HOME.
This mostly reverts 17af5d51de.
Suggested by Andy Wingo <wingo@igalia.com>.
* guix/ssh.scm (remote-daemon-channel)[redirect]: Remove 'read!' FFI
hack. Use buffered ports.
This fixes a regression in 'retrieve-files*' introduced in
896fec476f, whereby (guix scripts offload)
would not read the initial sexp now sent by the remote host via
'store-export-channel'. This would effectively prevent file retrieval
entirely when offloading.
* guix/ssh.scm (retrieve-files*): New procedure, like former
'retrieve-files' but with an extra #:import parameter.
(retrieve-files): Rewrite in terms of 'retrieve-files*'.
(file-retrieval-port): Make private.
* guix/scripts/offload.scm (transfer-and-offload): Pass #:import to
'retrieve-files*'.
(retrieve-files*): Remove.
Previously the redirect code could end up exiting prematurely because of
an uninteresting "exceptional condition" on the socket (info "(libc)
Waiting for I/O").
* guix/ssh.scm (remote-daemon-channel): Pass the empty list as the third
argument to 'select'. It was a mistake to pass a non-empty list there
in the first place.
Previously 'process-stderr' would always pass a bytevector of MAX-LEN to
then daemon in the %stderr-read case (i.e., 'import-paths'), instead of
LEN (where LEN <= MAX-LEN).
In practice the extra bytes didn't cause a protocol violation or
anything because they happen at the end of the stream, which typically
contains the canonical sexp of the signature, and the extra zeros were
just ignored.
* guix/serialization.scm (write-bytevector): Add optional 'l' parameter
and honor it.
* guix/store.scm (process-stderr): Pass LEN to 'write-bytevector'.
This works around <https://bugs.gnu.org/30066> and noticeably improves
performance when using GUIX_DAEMON_SOCKET=ssh://HOST (the redirect code
was transferring data to guix-daemon one byte at a time!).
* guix/ssh.scm (remote-daemon-channel)[redirect]: Define 'read!' and use
it instead of 'get-bytevector-some'.
This was harmless but non-compliant and unnecessary.
* guix/scripts/publish.scm <top level>: Rename "Guix-Nar-Compression" to
"X-Nar-Compression" as should have always been.
(render-nar, nar-response-port): Adjust accordingly.
(strip-headers): New procedure.
(sans-content-length, with-content-length): Use it.
Reported by Fis Trivial <ybbs.daans@hotmail.com>.
Fixes <https://bugs.gnu.org/28987>.
* guix/import/crate.scm (crate-fetch): Check whether the "license" info
is present.
This adjusts the workaround for <http://bugs.gnu.org/21093> so that it's
not limited to a single content-type.
* guix/scripts/publish.scm (render-nar/cached): Add the 'x-raw-file'
header on the response.
(render-content-addressed-file): Likewise.
(with-content-length): Remove the 'x-raw-file' header.
(http-write): Instead of dispatching on 'application/octet-stream',
check whether 'x-raw-file' is set to determine whether to spawn a
thread.
'guix copy --from' now reports messages much more useful than "failed to
retrieve files".
* guix/ssh.scm (store-export-channel)[export]: Wrap 'use-modules' in
'catch' and 'with-store' in 'guard'. Check for invalid items. Write a
status sexp on stdout.
(raise-error): New macro.
(retrieve-files): Read the initial status sexp and report errors
accordingly.
This saves 18 bytes on each 404 narinfo response.
* guix/scripts/publish.scm (render-narinfo): Pass #:phrase to
'not-found'.
(render-narinfo/cached): Likewise.
* guix/scripts/build.scm (numeric-extension?, tarball-base-name): New
procedures, formerly in 'package-with-source'.
(transform-package-source)[new-sources]: Look for '=' in URI. Each
element of the list of now a (PKG VERSION SOURCE) tuple.
Pass VERSION to 'package-with-source'.
(package-with-source): Add 'version' parameter and honor it.
* tests/scripts-build.scm ("options->transformation, with-source, PKG=URI")
("options->transformation, with-source, PKG@VER=URI"): New tests.
* doc/guix.texi (Package Transformation Options): Document the new
forms.
This is a followup to 614fffe427, which
broke cross-compilation to i686-w64-mingw32.
* guix/build-system/gnu.scm (standard-cross-packages): Add
"cross-libc:static" only when LIBC has a "static" output.
Fixes a regression whereby the host libc.a would be missing when
cross-compiling.
This is a followup to commit 6dff905e51.
* guix/build-system/gnu.scm (standard-cross-packages): Add the
"cross-libc:static".
This would crash 'guix refresh -u' with source URLs ending in
"…/v1.2.3".
* guix/upstream.scm (package-update): Be stricter when determining
ARCHIVE-TYPE.
Marius Bakke