Re: Support for MinGW compilers in the Cygwin environment

[Kai Tietz]

Hello,

Once again ...  We already discussed this issue some time ago directly, Earnie. 
 And just by repeating and spreading this nonsense over and over again, it 
doesn't make it right.  

First I have to state that therr is (and was) no target using *mingw64* as 
host-part triplet.  What should this host be?

A host using 64-bit architecture?  That doesn't make sense, due this part is 
already covered by the architecture-triplet-part.  This makes for Itanium 
ia64-*-*, or for amd64 x86_64-*-*.  This is well established naming schema used 
for toolchains.  Eg for Linux we name 32-bit target as i686-unknown-linux-gnu.  
For 64-bit Linux the triplet is x86_64-unknown-linux-gnu.  As you see no 64 in 
the host part ... it remains Linux, and the same is true for Windows targets 
... it remains Windows and its API is still the same.

I want to mention that there was already such a discussion on gcc back in 2008 
AFAIR.  At this time, there was the initial idea to name the 64-bit Windows 
target as mingw64 (and this seems to be still the status of Earnie's thinking) 
by me.  It was disliked by the communities of binutils/gdb/gcc by the given 
reason.  The bitness of the target is already specified by the 
architecture-part of the triplet.  There is no host called "mingw64" (as there 
is no host name *mingw32*). It was seen upstream as failure to agree on the - 
existing - *mingw32* in the past.  And so 64-bit target shares the same 
host-part as for 32-bit on Windows native target.  Secondly it was stated that 
the host-part isn't the name of a venture, instead it is the name of host.

I share both of these opinions.  There is already a Windows 64-bit target 
present for years, which shares host-part with 32-bit.  And from perspective of 
a binutils/gcc-maintainer for Windows targets I seriously decline the approach 
of introduce a 'mingw64' host part-triplet in config.guess.

So instead, we should change the use of *-*-mingw32* to *-*-mingw* instead.  
Well, we could agree on that even the name *mingw* is something wrong, and 
should something more host-specific as host-part of triplet.  But well, this is 
a different discussion.

Regards,
Kai