Page 1 of 1

DaemonR's weird toolchain

Posted: Sat Aug 30, 2014 9:20 pm
by Wajideu
There's a possible bug I wanted to point out with the GCC, but didn't think was important enough to create a new topic.

Foreword, I do have a rather odd toolchain layout. I'm using is Qt's i686-w64-mingw32 build and my Msys is located in the same directory as my GitHub installation.

But on to the actual issue, if you configure gcc-4.9.1 with the options

Code: Select all

--target=mips64 --enable-64-bit-bfd --enable-languages=c,c++ --disable-nls --disable-werror --disable-threads --with-arch=vr4300 --with-endian=big,little
It'll crash right towards the end of the build process for all-gcc with the error:

Code: Select all

../../src/gcc-4.9.1/gcc/hwint.h:109:2: error: #error Please add support for HOST_HALF_WIDE_INT
After a strenuous google search, it appears no one else has had this bug before. It's probably just my own dumbassery for having such a weird toolchain setup.

Regardless, I'm starting to get fed up with the GCC. Having to wait an hour to two hours for a toolchain to build, only for it to crash near the end and having to start all over because I was missing a single option or something is just aggravating as hell. Plus I think it's kinda ridiculous that there are no 8086/i186/i286 backends. Trying to add your own backend or object file format to the bfd in binutils is pure masochism. I'm strongly considering writing my own compiler at this point.

Re: GCC is broken

Posted: Sun Aug 31, 2014 1:12 am
by alexfru
DaemonR wrote:Regardless, I'm starting to get fed up with the GCC. ... Plus I think it's kinda ridiculous that there are no 8086/i186/i286 backends. Trying to add your own backend or object file format to the bfd in binutils is pure masochism. I'm strongly considering writing my own compiler at this point.
I did: Smaller C. :)

Re: DaemonR's weird toolchain

Posted: Sun Aug 31, 2014 10:26 am
by sortie
That's not a crash. That's a build error.

It likely happens because you do odd things. Stop doing that.

Re: DaemonR's weird toolchain

Posted: Sun Aug 31, 2014 2:15 pm
by Wajideu
alexfru wrote:I did: Smaller C. :)
That's a frontend, not a backend. My bad, thought you were talking about modifying the GCC itself.

But yeah, the main reason I use this layout is to save time and space. Kinda running low on memory at the moment so I'm being conservative. There's no reason really to have 2 toolchains and 2 shell environments that have the same target platform anyway.

I'm thinking that it may just be a missing header or something, so I'm going to try to install MinGW's headers over the top of my Qt toolchain directory later and see if that fixes it.