WIP: Build stage 2 GCC with plugin support and other minor GCC build improvements #168
Conversation
| @@ -45,6 +45,7 @@ if [[ "$RUN_CONFIG" = 1 ]] || [[ ! -f "$GCC_BUILD_PATH/Makefile" ]]; then | |||
| ;; | |||
| *mingw*) | |||
| TARGET_OPTIONS="$TARGET_OPTIONS \ | |||
| --libexecdir=$TOOLCHAIN_PATH/lib \ | |||
There was a problem hiding this comment.
This option is present in stage 2 build. Although there is no imminent issue without this option, it make thes result of stage 1 build consistent with the stage 2.
| @@ -59,6 +59,11 @@ if [[ "$RUN_CONFIG" = 1 ]] || [[ ! -f "$GCC_BUILD_PATH/Makefile" ]]; then | |||
| --without-libintl-prefix" | |||
| ;; | |||
| *mingw*) | |||
| # ADDED: --enable-plugin | |||
| # REMOVED: --with-gmp=$TOOLCHAIN_PATH | |||
There was a problem hiding this comment.
The REMOVED comments makes explicit what is the difference between this configuration and the upstream MSYS2 package configuration.
Blackhex
changed the title
| @@ -69,6 +74,7 @@ if [[ "$RUN_CONFIG" = 1 ]] || [[ ! -f "$GCC_BUILD_PATH/Makefile" ]]; then | |||
| --enable-cloog-backend=isl \ | |||
| --enable-version-specific-runtime-libs \ | |||
| --enable-lto \ | |||
| --enable-plugin \ | |||
There was a problem hiding this comment.
Should not it be enabled by default?
There was a problem hiding this comment.
For Linux target yes, for MinGW target is must be enabled explicitly as this feature is not used much accros the community. JFYI, MSYS2 cross-compiler packages do not enable this, the MSYS2 native MinGW packages does. That's why I've added the ADDED comment as I am synchronizing this configuration with the cross-compilation package for now. We may switch to synchronization with the native packages once we will succeed to build them.
There was a problem hiding this comment.
I would not change it here. I prefer to leave a comment about requirements in MSYS2 pipeline.
There was a problem hiding this comment.
Let me double check whether there is a difference between number of passing tests with and without this option enabled for cross-compiler testing. If this improves the number of passing tests, I'd keep it enabled for all targets. If it has no effect on the cross-compiler testing results, I'd enable it only for MSYS2 testing. Is this OK for you?
Blackhex
changed the title
This PR actually contains three separate changes, see in-code comments. IMO it would be overkill to submit it separatelly, but I can do it if needed.