Fix failures on MacOS #617
Conversation
We need ffi_closure_alloc() since commit cf10854 but don't use ffi_prep_closure() any longer.
When have_header('ffi.h') was successful the system libffi include
path is added to INCFLAGS. But when have_func() fails later on
the bundled libffi which should be used instead.
Using the wrong ffi.h resulted in compiler warnings about undefined
functions and leaded to segfaults on MacOS.
|
@tduehr Thank you for reviewing! IMHO it would be better to use a merge commit instead of squashing the commits the next time. It otherwise breaks the relation between the commit message and the source changes. If the commits belong semantically together, I do the squashing before PR'ing. Thank you! |
|
Makes sense, will do from here out. I normally squash stuff cause I'm a sucker for a linear, concise history. |
|
Ok, good to know. In contrast, having been forced to linear history by CVS and SVN until 10 years ago, I really enjoy the offline, octopus merge and rebase capabilities of git. |
|
I know a couple devs that manufacture a nonlinear history. Drives me crazy. My own repos normally have fluid histories near HEAD. I’m working on a PR I must have rebased my own history 20 times by now. Otherwise, my histories look like Charlie Brown’s xmas tree without the bend. |
Fix failures on MacOS intruduced by #540.
See the commit messages for detailed descriptions.