--- Begin Message ---
|
Subject: |
[PATCH 0/7] Lzip support for 'guix publish' and 'guix substitute' |
|
Date: |
Fri, 24 May 2019 15:31:59 +0200 |
Hi!
As a followup to Pierre’s work on (guix lzlib), these patches implement
‘lzip’ support for ‘guix substitute’ and ‘guix publish’. With these,
you can now run:
./pre-inst-env guix publish -Clzip …
on one side, and on another machine:
./pre-inst-env guix-daemon --build-users-group=guixbuild
and from there the client machine should be able to fetch
lzip-compressed substitutes.
These patches do not address the transitioning issue that we discussed
earlier, where we have clients lacking lzip support talking to an
lzip-capable server. As discussed earlier, clients will have to send
a special HTTP header, ‘X-Guix-Accept-Encoding’.
For the server-side, I’m still hesitating between implementing it in ‘guix
publish’ or simply running two instances of ‘guix publish’ side-by-side
(one gzip and one lzip) and letting nginx dispatch between the two. :-)
Note that we’ll have to adjust our nginx mirror configs to take that
header into account!
Comments?
Ludo’.
Ludovic Courtès (7):
lzlib: Add 'make-lzip-input-port/compressed'.
utils: Test 'compressed-port' and 'decompressed-port' for both gzip
and xz.
utils: Support compression and decompression with lzip.
publish: Add support for lzip.
self: Add dependency on lzlib.
gnu: guix: Add dependency on lzlib.
lzlib: 'lzread!' never returns more than it was asked for.
.dir-locals.el | 2 +
doc/guix.texi | 25 +++++--
gnu/packages/package-management.scm | 1 +
guix/lzlib.scm | 101 ++++++++++++++++++++++------
guix/scripts/publish.scm | 84 +++++++++++++++++------
guix/self.scm | 13 +++-
guix/tests.scm | 1 +
guix/utils.scm | 27 ++++++--
tests/lzlib.scm | 10 +++
tests/publish.scm | 36 ++++++++++
tests/utils.scm | 62 +++++++++++------
11 files changed, 284 insertions(+), 78 deletions(-)
--
2.21.0
--- End Message ---
--- Begin Message ---
|
Subject: |
Re: [bug#35880] [PATCH 1/7] lzlib: Add 'make-lzip-input-port/compressed'. |
|
Date: |
Mon, 27 May 2019 22:53:11 +0200 |
|
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/26.2 (gnu/linux) |
Hi!
Pierre Neidhardt <address@hidden> skribis:
> That said, if the encoder buffer is not empty, I think lz-compress-read
> should always return something >0.
Yes, probably. The docstring for ‘lz-compress-read’ says:
"Read up to COUNT bytes from the encoder stream, storing the results in
LZFILE-BV.
Return the number of uncompressed bytes written, a strictly positive integer."
^~~~~~~~~~~~~~~~~
However, the lzlib manual doesn’t say that for ‘LZ_compress_read’ (info
"(lzlib) Compression functions").
But that’s OK: the ‘read!’ method in ‘make-lzip-input-port/compressed’
can just call ‘lzwrite!’ again with more data when that happens, so I’ve
done that.
And I pushed the whole thing! :-)
I think it’d be good to let people play with it in their personal
setups.
Next up: multi-compression support in ‘guix publish’ (possibly?) so we
can smoothly transition on our build farms.
Thanks!
Ludo’.
--- End Message ---