To open-connection and port->connection. This overrides the discovered
builtin builders that the daemon says it provides.
This is useful when you want to generate compatible derivations that can be
run with a daemon that potentially doesn't support builtin builders that the
daemon you're using to generate the derivations has.
I'm looking at this in particular because I want to use this in the data
service, since it provides substitutes for derivations, and since these can be
built on other machines, it's useful to control which builtin builders they
depend on.
* guix/store.scm (open-connection, port->connection): Accept
#:built-in-builders and use this instead of %built-in-builders.
Fixes: <https://issues.guix.gnu.org/67250>.
Change-Id: I45d58ab93b6d276d280552858fc81ebc2b58828a
Remove the inner connect procedure, as now that #:non-blocking? needs passing
on, this just makes things more difficult. This commit also fixes not passing
#:non-blocking? on in the case where open-unix-domain-socket is called as
connect.
* guix/store.scm (connect-to-daemon): Refactor and fix non-blocking
connections to sockets with a filename.
Change-Id: I61cd99920df91baba95567d670bec6fa94043875
For some applications, it's important to establish a non-blocking connection
rather than just making the socket non-blocking after the connection is
established. This is because there is I/O on the socket that will block during
the handshake.
I've noticed this blocking during the handshake causing issues in the build
coordinator for example.
This commit adds a new with-store variant to avoid changing the behaviour of
with-store/open-connection to ensure that this change can't break anything
that depends on the blocking nature of the socket.
* guix/store.scm (open-unix-domain-socket, open-inet-socket): Take
#:non-blocking? and use SOCK_NONBLOCK when calling socket if appropriate.
(connect-to-daemon, open-connection, call-with-store): Take #:non-blocking?
and pass it on.
(with-store/non-blocking): New syntax rule.
* .dir-locals.el (scheme-mode): Add entry for with-store/non-blocking.
Change-Id: I8225762b78448bc1f7b698c8de5d736e13f577bf
(web …) modules pull in (gnutls) indirectly. Arrange to load them
lazily, thereby reducing I/O and allocations when GnuTLS is not needed
such as when running ‘guix describe’ or ‘guix shell’ on a cache hit.
* guix/download.scm: Autoload (web uri).
* guix/scripts/describe.scm: Likewise.
* guix/store.scm: Likewise.
(%default-substitute-urls): Remove ‘resolve-interface’ call and use
https URLs unconditionally.
Change-Id: Ide470c556a14866e8740966d25821df487a79859
The aim here is to improve the user experience. There's anecdotal evidence
that the network performance for bordeaux is better compared to ci at least
for some users, and I don't know of any issues with rate limiting or access
restriction for bordeaux compared to ci. It also has IPv6 support.
Additionally, bordeaux generally had more substitutes than ci, particularly
for aarch64-linux and armhf-linux. This change will offer a very slight
speedup for those substitutes that only bordeaux has.
Bordeaux has been a default substitute server for nearly 3 years now and I
think this change is overdue. I'm also hopeful that we'll be able to build on
the testing regarding mirrors for bordeaux, and that'll allow potentially
improving the hosting setup (through providing more redundancy) and further
improving substitute fetching for users who currently have issues with
substitute access.
* config-daemon.ac: Switch substitute urls order.
* doc/guix.texi: Ditto.
* etc/guix-install.sh: Ditto.
* gnu/installer/newt/network.scm (wait-service-online): Ditto.
* guix/store.scm (%default-substitute-urls): Ditto.
Change-Id: I4f6d93ae1fc8b03d80b47b18b5749a51f1fde17b
Signed-off-by: Christopher Baines <mail@cbaines.net>
This was obtained by setting up this environment:
guix shell -D guix --with-input=guile@3.0.9=guile-next \
--with-commit=guile-next=e2ed33ef0445c867fe56c247054aa67e834861f2
-- make -j5
then adding 'unused-module' to (@@ (guix build compiler) %warnings),
building, and checking all the "unused module" warnings and removing
those that were definitely unused.
The goal is to allow (guix grafts) to use (guix gexp) without
introducing a cycle between these two modules.
* guix/grafts.scm (%graft?, call-without-grafting, without-grafting)
(set-grafting, grafting?): Move to...
* guix/store.scm: ... here.
Previously, 'guix shell' for example would leak the socket that's
connected to the daemon.
* guix/store.scm (open-unix-domain-socket, open-inet-socket): Pass
SOCK_CLOEXEC to 'socket'.
* tests/guix-shell.sh: Add test.
Fixes <https://issues.guix.gnu.org/55721>.
Starting from the switch to functional object caches ca.
9e5812ac59, we would be losing accumulated
caches when aborting to the build handler. This patch fixes that.
In particular, by preserving '%reference-cache-id', we avoid redundant
'query-references' RPCs, which accounted for a large part of the extra
processing time.
* guix/store.scm (build-accumulator): When returning an <unresolved>
node, call 'set-store-connection-caches!' before and after to preserve
caches.
(map/accumulate-builds): Pass STORE as the first argument to
the <unresolved> continuation.
This reduces the wall-clock time of:
./pre-inst-env guix system vm gnu/system/examples/desktop.tmpl -n
from 2m13s to 53s (the timings depend on which derivations have already
been built and are in store; in this case, many were missing).
* guix/store.scm (default-cutoff): New variable.
(map/accumulate-builds): Use it. Parameterize it in recursive calls to
have decaying cutoff.
Users of 'mcached' can now specify a cache ID; furthermore, the cache
hit rate is automatically recorded for all the caches accessed with
'mcached'.
* guix/store.scm (%max-store-connection-caches)
(%store-connection-cache-names): New variables.
(recorder-for-cache): New procedure.
(record-cache-lookup!): Add 'cache-id' parameter and rewrite in terms of
'recorder-for-cache'.
(lookup-cached-object): Add 'cache-id' parameter and honor it.
(%mcached): Add #:cache parameter and honor it.
(mcached): Add '=>' keyword and corresponding clauses.
* guix/deprecation.scm (warn-about-old-daemon): Mention a direct consequence
of failing to upgrade the daemon.
* guix/store.scm (build-things): Add a comment explaining the rationale to
emit a deprecation warning for a daemon older than version 0x163.
Suggested-by: Ricardo Wurmus <rekado@elephly.net>
Fixes <https://issues.guix.gnu.org/51732>.
Regression introduced in 2015d3f042.
The (eq? store expected-store) comparison doesn't work on this branch
where the functional cache, with 'cache-object-mapping', is used
extensively, thereby changing the store's object identity.
* guix/store.scm (build-accumulator): Compare the socket of STORE and
EXPECTED-STORE rather than the store themselves.
This is a followup to b19250eec6,
providing a proper fix for <https://issues.guix.gnu.org/46756>.
* guix/remote.scm (remote-eval): Revert b19250eec6.
* guix/store.scm (build-accumulator): Turn into a procedure. Call
CONTINUE when the store is not eq? to the initial store.
(map/accumulate-builds): Adjust accordingly.
* tests/store.scm ("map/accumulate-builds and different store"): New test.
Fixes <https://issues.guix.gnu.org/50264>.
Reported by Lars-Dominik Braun <lars@6xq.net>.
This fixes a regression introduced in
fa81971cba whereby 'map/accumulate-builds'
would return REST (the tail of LST) without applying PROC on it. The
effect would be that 'lower-inputs' in (guix gexp) would dismiss those
elements, leading to derivations with correct builders but only a subset
of the inputs they should have had.
* guix/store.scm (map/accumulate-builds): Add #:cutoff parameter and
remove 'accumulation-cutoff' variable. Call PROC on the elements of
REST.
* tests/store.scm ("map/accumulate-builds cutoff"): New test.
Fixes <https://bugs.gnu.org/49439>.
Reported by Ricardo Wurmus <rekado@elephly.net>.
Previously, a command such as:
guix environment pigx-scrnaseq
could lead to unbounded memory growth and could even fail to complete
when some items are missing from the store. This was because
'map/accumulate-builds' callees would keep making .drv build requests
that were turned into <unresolved> nodes; in this case, there are often
many identical build requests. Stopping accumulation earlier allows us
to unlock the situation by proceeding with the first few build requests
instead of spinning until we've accumulated all the build requests.
* guix/store.scm (map/accumulate-builds): Define 'accumulation-cutoff'.
Use a loop when iterating over LST and maintain a counter of unresolved
nodes met so far; return when the counter exceeds ACCUMULATION-CUTOFF.
In addition to substitutes from ci.guix.gnu.org. There are more changes that
can be made in the future, but these changes seem like a good start.
* config-daemon.ac (guix_substitute_urls): Add https://bordeaux.guix.gnu.org.
* guix/scripts/substitute.scm (%default-substitute-urls): Add
http://bordeaux.guix.gnu.org.
* guix/store.scm (%default-substitute-urls): Add bordeaux.guix.gnu.org.
* doc/guix.texi: Adjust accordingly.
* doc/contributing.texi: Adjust accordingly.
This procedure lost its only user in commit
710854304b.
* guix/store.scm (references/substitutes): Remove.
* tests/store.scm ("references/substitutes missing reference info")
("references/substitutes with substitute info"): Remove.
This avoids a closure allocation when 'lookup-cached-object' is called.
* guix/store.scm (lookup-cached-object): Avoid optional/keyword
arguments and inline.
(%mcached): Adjust accordingly.
The "store-path-hash-part #f", due to a SRFI-64 bug, was marked as
successful even though 'store-path-hash-part' was throwing an exception.
* guix/store.scm (store-path-hash-part): Really return #f.
* guix/serialization.scm (read-string-pairs): New procedure.
* guix/store.scm (read-arg): Add support for 'string-pairs'.
(find-roots): New procedure.
* tests/store.scm ("add-indirect-root and find-roots"): New test.
Fixes <https://bugs.gnu.org/42047>.
Reported by Jan Nieuwenhuizen <janneke@gnu.org>.
* guix/store.scm (open-connection): Use '=' instead of 'eqv?'.
This works around <https://bugs.gnu.org/42060> while also being more
accurate since the arguments are known to be integers.
This ensures the stack is not unwound before the exception is re-thrown,
as was the case since 8ed597f4a2, leading
to '&store-protocol-error' being uncaught by 'with-error-handling'
in (guix scripts build) & co.
* guix/store.scm (call-with-store): Define 'thunk'. Add 'cond-expand'
to use 'with-exception-handler' on 'guile-3' and 'catch' otherwise.
Fixes <https://bugs.gnu.org/40428>.
Reported by Marius Bakke <mbakke@fastmail.com> and 白い熊.
Regression introduced with the first uses of 'with-build-handler' in
commit 62195b9a8f and subsequent.
* guix/store.scm (call-with-store): Use 'catch #t' instead of
'dynamic-wind'. This ensures STORE remains open when a non-local exit
other than an exception occurs, such as an abort to the build handler
prompt.
* tests/store.scm ("with-build-handler + with-store"): New test.
* guix/store.scm (<unresolved>): New record type.
(build-accumulator, map/accumulate-builds, mapm/accumulate-builds): New
procedures.
* tests/store.scm ("map/accumulate-builds", "mapm/accumulate-builds"):
New tests.
* guix/store.scm (current-build-prompt): New variable.
(call-with-build-handler, invoke-build-handler): New procedures.
(with-build-handler): New macro.
* tests/store.scm ("with-build-handler"): New test.
