On 4/6/24 5:54 PM, Bruno Haible wrote:
> It was caused by empty directories, which cannot be committed into git.
> Thus in your checkout these directories are gone, while in my work
> directory, they were still present.
Ah, that was my second guess. It works now, thanks.
Collin
Hi Collin,
> Some of the gettext tests fail with GNULIB_TOOL_IMPL=sh due to
> "Creating directory ..." messages. I ran into this while looking into
> a separate issue with gnulib-tool.py.
>
> Here is a list of them:
>
> ./test-gettext-1.sh
> +Creating directory gettext-runtime/doc
>
> ./test-ge
On 4/6/24 9:37 AM, Bruno Haible wrote:
> OK, I have added it to the test suite now. (Also because one of its
> gnulib-tool
> invocations uses two --local-dir options — which we haven't had in the test
> suite so far.)
Some of the gettext tests fail with GNULIB_TOOL_IMPL=sh due to
"Creating direct
Hi Bruno,
On 4/6/24 9:37 AM, Bruno Haible wrote:
> OK, I have added it to the test suite now. (Also because one of its
> gnulib-tool
> invocations uses two --local-dir options — which we haven't had in the test
> suite so far.)
Thanks! Yes, I remember you had to explain how gettext uses
--local-
Hi Bruno,
On 4/6/24 6:41 AM, Bruno Haible wrote:
> * gnulib-tool.py is between 2x and 300x faster than gnulib-tool.sh, depending
> on the arguments. This means, the goal of speedup has been fully achieved.
> The other goals (clear code, maintainability, etc.) are still present.
Yes, I meant t
Collin Funk wrote:
> It seems that gettext catches a lot of issues with gnulib-tool.py...
OK, I have added it to the test suite now. (Also because one of its gnulib-tool
invocations uses two --local-dir options — which we haven't had in the test
suite so far.)
I'll reply to your patches later.
B
Hi Collin,
> I was looking at simplifying some list comprehensions yesterday and
> noticed that using sets + removing unnecessary sorted() calls in
> GLModuleTable reduces the time of ./import-tests/test-all.sh from 4.8
> seconds to 4.5 seconds consistently. A truly life changing amount of
> time.
Hi Bruno,
On 4/6/24 6:13 AM, Bruno Haible wrote:
> Done. I added a 'test-pspp-1.sh' test.
Thanks! It passes for me.
Collin
> Done. I added a 'test-pspp-1.sh' test.
On this test, the Python implementation is ca. 80x faster than the shell
implementation:
$ GNULIB_TOOL_IMPL=py time ./test-pspp-1.sh
real 0.43s
user 0.33s
sys 0.09s
$ GNULIB_TOOL_IMPL=sh time ./test-pspp-1.sh
real 34.52s
user 36.66s
sys 6.73s
It seems that gettext catches a lot of issues with gnulib-tool.py...
The first patch addresses this:
diff -ru
/home/collin/.local/src/gettext/gettext-runtime/gnulib-m4/gnulib-cache.m4
/home/collin/.local/src/glpyVCpuH3/gettext-runtime/gnulib-m4/gnulib-cache.m4
--- /home/collin/.local/src/gettex
On 2024-04-04 I wrote:
> I think I will add a unit test for this one.
Done. I added a 'test-pspp-1.sh' test.
Bruno
When running ./autogen.sh in gettext gnulib-tool.py fails on the first
invocation. :(
The command is something like this:
$GNULIB_TOOL --dir=gettext-runtime ...
We get the following error message:
Traceback (most recent call last):
File "/home/collin/.local/src/gnulib/.gnulib-tool.py", li
On NetBSD 10.0/i386 I see this test failure:
FAIL: test-expm1l
=
../../gltests/test-expm1.h:41: assertion 'y >= x' failed
FAIL test-expm1l (exit status: 134)
The cause is that for arguments such as x = 2^-1075, expm1l(x) returns 0,
which it never should. (exp(x) - 1 is always ≥ x
13 matches
Mail list logo
