On Wed, Jan 15, 2020 at 02:22:45PM +0900, Kyotaro Horiguchi wrote:
> At Tue, 14 Jan 2020 22:34:10 +0100, Peter Eisentraut <peter.eisentraut@2ndquadrant.com> wrote in
>> On 2020-01-09 10:56, Peter Eisentraut wrote:
>>> Note that there is FILEFLAGSMASK and FILEFLAGS. The first is just a
>>> mask that says which bits in the second are valid. Since both
>>> libpq.rc
>>> and win32ver.rc use FILEFLAGS 0, it doesn't matter what we set
>>> FILEFLAGSMASK to. But currently libpq.rc uses 0x3fL and win32ver.rc
>>> uses 0x17L, so in order to unify this sensibly I looked for a
>>> well-recognized standard value, which led to VS_FFI_FILEFLAGSMASK.
Hmm. I agree that what you have here is sensible. I am wondering if
it would be better to have VS_FF_DEBUG set dynamically in FILEFLAGS in
the future though. But that's no material for this patch.
>> Here is a rebased patch.
>
> It applied on 4d8a8d0c73 cleanly and built successfully by VS2019.
I have been testing and checking the patch a bit more seriously, and
the information gets generated correctly for dlls and exe files. The
rest of the changes look fine to me. For src/makefiles/Makefile.win32,
I don't have a MinGW environment at hand so I have not directly
tested but the logic looks fine.
--
Michael
