[nanomsg] Re: Problem with x64 build on windows
- From: Martin Sustrik <sustrik@xxxxxxxxxx>
- To: nanomsg@xxxxxxxxxxxxx
- Date: Thu, 26 Sep 2013 11:11:27 +0200
On 26/09/13 11:07, Immanuel Weber wrote:
Hi Martin,
yesterday I allmost finished that modification. This morning a simpler
solution came to my mind: just include the efd.h header in device.c and
exchange the use of int for the file descriptors there from to nn_fd.
The patch for that is below, if you prefer your proposed solution, I can
provide a patch for that too.
Yes, please. There's no point in introducing a dependency between efd
and device. Let them rather both depend on nn_fd.
p.s. there are some compiler warnings about conversions scattered across
the project (regardless of the patch), where data loss can happen. Some
of them are quite severe, like size_t to uint8_t, which is on my machine
a 8 to 1 byte conversion. I will try have a look at some of them.
Right. I would say an assert and an explicit cast would work in such case:
size_t sz;
...
nn_assert (sz < 256);
uint8_t c = (uint8_t) sz;
Martin
Other related posts:
