| 20 Feb 2025 |
 emily | because you still end up with C++ files, maybe internal headers, whatever on the path | 19:05:27 |
 John Ericson | agreed | 19:05:44 |
| John Ericson | and I want private non-installed headers | 19:05:49 |
| emily | perhaps it would be best if the libraries were entirely separate Meson builds that consume each other via .pc files, but I don't know if Meson itself can orchestrate that or if you'd have to just do it with Nix. | 19:06:00 |
| John Ericson | which further motivates a separation | 19:06:01 |
| John Ericson | emily: we are trying to do the meson version of taht | 19:06:21 |
| John Ericson | * emily: we are trying to do the meson version of that | 19:06:23 |
| John Ericson | with meson subprojects | 19:06:27 |
| John Ericson | but it is designed to not require installing | 19:06:36 |
| emily | perhaps you can use fs.copyfile to copy headers into an include/nix/* directory in the build tree and then have that be the include directory | 19:06:50 |
| John Ericson | (installing is all side effects not ninja target direven, like most build systems) | 19:06:54 |
| John Ericson | fs.copyFile will do it at configure time | 19:07:02 |
| emily | I think that that should work and satisfy ^ strongly | 19:07:07 |
| emily | it says "Copy a file from the source directory to the build directory at build time", is that inaccurate? | 19:07:18 |
| John Ericson | which will kill the debug cycle | 19:07:19 |
| emily | it returns a custom_target() | 19:07:22 |
| emily | which would run at build time, afaik. | 19:07:27 |
| John Ericson | oh woah | 19:07:29 |
| John Ericson | ok sorry | 19:07:34 |
| John Ericson | I didn't know any of the fs.* things were deferred actions | 19:07:45 |
| emily | so you enumerate (or glob I guess if you have to) the headers from source – that part would be configure time, I suppose – fs.copyfile them to include/nix, and include in the build directory that can be the resulting library's include directory | 19:08:16 |
| John Ericson | we already don't have glob (RIP) | 19:08:26 |
| John Ericson | so making a bunch of separate build targets is OK | 19:08:35 |
| emily | seems like a silly amount of ceremony but you can probably abstract it into the existing support thing. | 19:10:33 |
| emily | and that at least ensures that downstream code only sees the include/ as it will be installed | 19:10:49 |
| emily | regardless of source organization | 19:10:54 |
| John Ericson | yes | 19:10:59 |
| John Ericson | emily: I did the symlink wrong and broke git :D | 19:22:37 |
| emily | still don't think you need an in-source symlink at all :P | 19:23:36 |
| John Ericson | I don't but I think it is an easier first step in some respects | 19:24:41 |
