Commit Graph
23 Commits
Author SHA1 Message Date
Carsten Hartenfels d89c01b39f Intuit if we should use anonymous structs
Rather than making the user #define something explicitly, we can guess
based on the compiler type and C standard.
2019-11-24 16:02:12 -05:00
Carsten Hartenfels b9aa14d25e Add a test for struct type initializers
To make sure that they all work properly and none of them trigger
warnings.
2019-11-24 16:02:12 -05:00
Carsten Hartenfels 8ad273475f Make GLMS_ initializers use the GLM_ versions
So that we don't write them twice. More consistent this way.
2019-11-24 16:02:03 -05:00
Carsten Hartenfels c8211b3a62 Re-order struct type contents array-first
So that initializers will prefer the array entry, rather than trying to
initialize the anonymous struct.
2019-11-24 16:01:55 -05:00
Carsten Hartenfels 11dae5126b Add braces around matrix struct intializers
They're missing and trigger warnings in the tests.
2019-11-24 13:32:35 -05:00
Carsten Hartenfels 390a5035a8 Use gnu11 instead of gnu99 to get rid of warnings
CGLM uses anonymous structs, which is a C11 feature. When trying to
build the tests in C99 mode, you get warnings to that effect. Switching
to C11 fixes this.
2019-11-24 13:32:35 -05:00
Carsten Hartenfels 267348af03 Build tests with the same flags as the library
In particular, with the same warning flags. That means it now warns
about a few things during compilation that the following commits are
gonna fix.
2019-11-24 13:32:35 -05:00
Carsten Hartenfels f9abf2a7df Don't add default autoconf flags to build
By default, it adds something like `-O2 -g` to the cflags. That
conflicts with the flags in Makefile.am, which specifies -O3. This
commit removes the default flags so we only get what we actually
specify.
2019-11-24 13:32:35 -05:00
Carsten Hartenfels 4b9b7aeb20 Add a vec2s struct type for consistency
There's a vec2 type, so there should probably be a struct version of it
too. Even if no functions use it right now, if a library user (like me)
needs a 2-element vector, they don't need to roll their own.
2019-11-23 14:37:48 -05:00
Carsten Hartenfels c83f25343f Error out on invalid empty prototypes
This way, a function prototype like `glms_mat3_identity()` will not
compile, instead you have to change it to the proper
`glms_mat3_identity(void)`.
2019-11-23 10:10:27 -05:00
Carsten Hartenfels f3ea5b4f3e Change empty prototypes to (void)
Because () means an arbitrary number of arguments in C, which is not
intended here.
2019-11-23 10:10:17 -05:00
Carsten Hartenfels a651827012 Make things work for ancient pkg-config versions
Travis CI uses a really, really old pkg-config apparently, which doesn't
have the PKG_INSTALLDIR macro. In that case, we have to emulate it
manually.
2019-06-23 15:12:21 +02:00
Carsten Hartenfels caba5b3c7c Document pkg-config usage and prefix pitfalls 2019-06-23 14:04:34 +02:00
Carsten Hartenfels b231645131 Capitalize Makefiles, sometimes matters on Linux
Lowercased Makefiles don't get picked up by the .gitignore for example,
which always looks odd after configuring. This commit just puts a
capital 'M' in front, like it's common.
2019-06-23 13:48:02 +02:00
Carsten Hartenfels 99cff9e74e Add pkg-config support 2019-06-23 13:47:09 +02:00
Carsten Hartenfels bbb52f352c Add missing struct headers to makefile.am 2019-06-23 13:24:06 +02:00
Carsten Hartenfels f7b45776e9 Find post-build.sh even when not building in .
You can make configure build in a different directory than the current
one when you give it a `--prefix` option. When doing that, the current
directy will be the build directory, not the source directory. This
breaks running `./post-build.sh`.

This commit fixes it by replacing the invocation with something that
takes into account `$VPATH` which properly references the source
directory. The `post-build.sh` still works properly because it
references the current directory to do its thing, which will (correctly)
be the build directory.
2019-06-01 14:14:58 +02:00
Carsten Hartenfels eefafefbae Also add a load of other number constants
These are all missing from standard C, so we might as well define em.

See https://github.com/recp/cglm/pull/61#discussion_r219406859
2018-09-21 20:03:28 +02:00
Carsten Hartenfels b9021978cb Replace the use of deprecated CGLM_PI* constants 2018-09-21 19:59:23 +02:00
Carsten Hartenfels 3fd12032e6 Clean up pi constants, deprecating the old names
GLM_PI* is now used for the double versions, GLM_PI*f for the float
versions. The CGLM_ prefixed versions are now deprecated, since that
prefix is kinda only used for constants.

See https://github.com/recp/cglm/pull/61#issuecomment-423069770
2018-09-21 07:29:54 +02:00
Carsten Hartenfels 59ee8c1fd2 Use even more precise definitions for pi constants
See https://github.com/recp/cglm/pull/61#issuecomment-422955122
2018-09-21 07:26:55 +02:00
Carsten Hartenfels b00f2b9ccc Replace M_PI_4 in test_affine with CGLM_PI_4
As in the previous commit, because it's non-standard and depending on
your settings it can fail to compile because of it.
2018-09-19 20:46:37 +02:00
Carsten Hartenfels d3c50147cb Replace non-standard M_PI* constants with literals
M_PI, M_PI_2 and M_PI_4 aren't part of the C standard. If you put your
gcc into strict standards mode, like `-std=c11`, you won't get these
constants and including cglm.h will fail.

This commit replaces those constants with their literal values. The cast
to a float remains, to keep exactly the same behavior as before.
2018-09-19 20:42:05 +02:00