Note that while upstream updated their licencing docs in this release
to say there's 2 licences, not 3, they do still say:
> Note that the libjpeg-turbo SIMD source code bears the
> [zlib License](https://opensource.org/licenses/Zlib), but in the context of
> the overall libjpeg API library, the terms of the zlib License are subsumed
> by the terms of the IJG License.
And by our handling of licences, that seems to me like we should still
list the zlib licence.
Signed-off-by: Sam James <sam@gentoo.org>
backport fix for round.m4, exit() -> return
Also fix bashism:
test $var = no
fails if $var is only ever defined after this line. gnulib is brilliant
software and naturally does exactly this, because it is highly logical
to only define code after you try to use it.
Probably. It's hard to tell because the actual code that gets run is:
func_gl_gnulib_m4code_5264294aa0a5557541b53c8c741f7f31
since the other sparkling aspect of gnulib's innovative brilliance is
the realization that *of course* the most robust way to design software
is to use checksums as *replacements* for function names.
Or, "that time Eli got jiatanned by gnulib".
The quoting was silently "fixed" upstream in:
https://github.com/coreutils/gnulib/commit/db1ee11e2168af7137db53289a92e306d2277b0b
It is the same "fix" we do here, thus restoring dash <--> bash
compatibility and preventing:
```
* QA Notice: Abnormal configure code
*
* ./configure: 38012: test: =: unexpected operator
```
although it seems to me gnulib is just more generally broken.
Closes: https://bugs.gentoo.org/900346
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
Signed-off-by: Sam James <sam@gentoo.org>
This release fixes an issue where under certain circumstances
(in particular when invoked within the Gentoo sandbox), `multi_wait()`
would incorrectly calculate the timeout, resulting in test failures.
Allow setting `--enable-debug` via USE=debug with an appropriate warning
to the ebuild and live ebuild.
Closes: https://bugs.gentoo.org/932660
See-also: https://github.com/curl/curl/issues/13782
Signed-off-by: Matt Jolly <kangie@gentoo.org>
If USE=-mpi, a file in src_test is NOT sedded to fix a bug in the file,
because the file does not exist. But this condition was incorrectly
coded.
In bash:
```
cmd1 && cmd2 || cmd3
```
is a code smell. If either of the first two commands fails, the third
command is run -- in this case, die. In other words, the first two
commands were *supposed* to be "only run cmd2 if it makes sense to".
Instead, if it "doesn't make sense to" run cmd2, the die was triggered.
A more general solution is to spec the build format to require all
commands to pass without manually `die`ing (leading to the use of &&
ceasing in general), but that is not how portage works. Either way, the
solution is using `if` as intended.
ref. https://mywiki.wooledge.org/BashGuide/TestsAndConditionals
Closes: https://bugs.gentoo.org/933136
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
Signed-off-by: Sam James <sam@gentoo.org>
Sam James
Eli Schwartz
Matt Jolly
Yixun Lan
Patrick McLean
Aliaksei Urbanski
Conrad Kostecki