On 2/3/26 05:36, Richard Henderson wrote:
On 1/20/26 00:05, Michael Tokarev wrote:
Since CONFIG_EPOLL is now unused, it's okay to
perform this rename, to make it less ugly.
Since epoll is linux-specific and is always present on linux,
define CONFIG_EPOLL to 1 on linux, without compile-time test.
Signed-off-by: Michael Tokarev <[email protected]>
---
meson.build | 3 +--
util/aio-posix.h | 4 ++--
util/meson.build | 2 +-
3 files changed, 4 insertions(+), 5 deletions(-)
diff --git a/meson.build b/meson.build
index f4eae50764..157189e306 100644
--- a/meson.build
+++ b/meson.build
@@ -2673,8 +2673,7 @@ config_host_data.set('CONFIG_INOTIFY1',
have_inotify_init1)
# has_header_symbol
config_host_data.set('CONFIG_BLKZONED',
cc.has_header_symbol('linux/blkzoned.h',
'BLKOPENZONE'))
-config_host_data.set('CONFIG_EPOLL_CREATE1',
- cc.has_header_symbol('sys/epoll.h',
'epoll_create1'))
+config_host_data.set('CONFIG_EPOLL', host_os == 'linux')
I guess my question remains: Why keep CONFIG_EPOLL if it is identical to
the existing CONFIG_LINUX?
It would be less obvious this way in the code. When you've some
#ifdef, when using CONFIG_EPOLL, it's much more obvious why it's here.
When it's CONFIG_LINUX, you start to wonder which of the 100500
different linux-only features this place is about.
Current code very obviously uses configuration test for particular
_feature_ (presence of epoll). Yes, this interface is linux-specific,
but in my view it is better to keep depending on particular feature,
than to check of the os type or library version or something else like
that.
Thanks,
/mjt
