From one machine at home I get a short fetch from ftp(1) (which is the
file I generated distinfo for, but it untarred sufficiently ok that I
didn't notice...) From another machine I get the full file, and from the
original machine with curl I get the full file(!)
Mirror the distfile as well, since there's something weird going on
somewhere.
Fetch issue reported by phessler.
The x86 assembly FFT implementation uses dispatch tables in .text. This is
incompatible with x-only so move them into .data.rel.ro where we can have
relocations. But since they are in the RELRO segment they will be made
read-only by ld.so after relocations have been processed.
incompatible with x-only so move them into .data.rel.ro where we can have
relocations. But since they are in the RELRO segment they will be made
read-only by ld.so after relocations have been processed.
ok robert@ (who tested i386)
pigz starts with "stdio rpath wpath cpath fattr chown" like base gzip(1).
For `g.pipeout' aka. -c --stdout and `g.decode == 2' aka. -t --test,
pigz drops to "stdio rpath cpath" while gzip uses "stdio rpath".
The original pledge diff added "cpath" for the case where pigz's signal
handler unlink(2)s output files, but this is actually not needed since
-c and -t never open any file for output/writing.
Drop "cpath" to match gzip behaviour.
Do the same for `g.list' aka. -l --list.
Since the riscv64 target name changed its position in the sorted list
needs to be adapted, else the target can't be found. No bump needed
since it didn't build, fix present in the riscv64 bulk build started on
Jan 7.
The configuration format changed! Existing configurations and themes
must be rewritten, although most parts are the same.
Other interesting changes:
- support for multiple accounts
- added a built-in editor
full changelog: https://github.com/RasmusLindroth/tut/releases
Since recent x-only changes, the bootstrap segfaults very early on in the
build. The obvious workarounds like USE_NOEXECONLY or passing in variants
of --no-execute-only via the environment aren't enough.
Institute. It has been developed specifically to increase legibility for
readers with low vision, and to improve comprehension.
Named for the founder of Braille Institute, Atkinson Hyperlegible is a
traditional grotesque sans-serif at its core. It departs from tradition
to incorporate umambiguous, distinctive and at times, unexpected always
with the goal of increasing character recognition, and ultimately
improve reading.
a bit of cleanup suggested by op@
OK op@ sthen@
- fix comparison in tree object parser always evaluating to false (jamsek)
- add missing bounds-check in gitproto ref-line tokenizer
- gotd.8 and gotwebd.8 man page improvements (patch by Josiah Frentsos)
- make gotd session process accept just one flush packet at a time
- sort ENVIRONMENT entries in got.1 and tog.1 (op)
- only forward implicit flush packets from gotsh if they are expected
- return GOT_ERR_EOF from pkt.c if a read attempt indicates EOF
- treat read errors from client socket in gotsh as fatal errors
- gotwebd: refactor gotweb_render_content_type/_file (op)
- gotwebd: turn gotweb_get_time_str into gotweb_render_age (op)
- gotwebd: don't list references per-commit in got_get_repo_commits (op)
- gotwebd: avoid history traversal in briefs/commits, kill "prev" button (op)
- gotd: disconnect on client EOF error to avoid stale connections (op)
- gotsh: validate with parse_command before connecting (op)
- add a test for ssh connections to gotsh without a repo path argument (op)
- gotsh: avoid a temporary buffer for the socket path (op)
- gotsh: move apply_unveil right after the first pledge (op)
- fix tog refreshing the screen more often than intended (jamsek)
- tog: show action report on user-toggleable event (patch by Mikhail, jamsek)
- don't print empty line when exiting tog (jamsek)
- got: don't leak pathlist in commit and revert commands (jamsek)
- gotd: fix wrong PIDs for parent and child processes shown in logs