On Wed, Jan 13, 2021 at 08:20:34PM +0300, Dmitry Kozlyuk wrote:
>       pci/windows: fix build with SDK 10.0.xxxxx
> 
>       NetUIO device class and interface GUIDs are defined in system
>       headers starting from platform SDK v10.0.xxxxx. Inspect SDK version
>       to avoid redefinition.
> 
>       Fixes: hhhhhhhhhhhh ("...")
>       Cc: sta...@dpdk.org
> 
>       Signed-off-by: Tyler Retzlaff <roret...@linux.microsoft.com>
> 
> I can't find which "xxxxx" corresponds to NTDDI_WIN10_FE.

there won't be a specific sdk/kit that corresponds. the best i can offer
is to include in the message the latest sdk/kit version for that platform
version.

also be aware since there is no 1:1 alignment between NTDDI_VERSION and
the sdk/kit build where the conflicting GUID definitions were introduced
there will still be a subset of eal/pre-release kits if installed that will
lead to compilation failure of dpdk even with this change. i'll also add
a note in the commit message about this limitation.

windows doesn't promise api compatibility for pre-release kits and there is
no approved mechanism / versioning to provide such compatibility.

> "hhhhhhhhhhhh" are 12 chars of the hash of the commit which introduced the
> code you fixed. "..." is said commit subject. This info is needed to backport
> your fix to stable branch, so that it can also be built with newer SDK. May
> not be very important until Windows port gets its users, I mostly tell it to
> illustrate the process.

yes, understood this one is definitely appropriate for backport

> > +#if (! defined(NTDDI_WIN10_FE) || NTDDI_VERSION < NTDDI_WIN10_FE)
> 
> Braces are redundant here and DPDK style is to avoid them (see rte_common.h).

yes, sorry will fix. best part is i don't like redundant braces i just added
them because i encounter so many people who do. (can't win really)

Reply via email to