Well, debuild seem to put the same CFLAGS/LDFLAGS to gcc-mingw-w64 as it does to gcc, and i guess one of the "debian hardening" options puts some flags to mingw's LD that does not fly. This does not happen when you build "manually", so it is quite likely just a build config thing.
Sorry, for this noise, as this is most certainly unrelated to the original post. I will do some investigations towards debian build settings, but i do not think it has much to do with the gcc-11 concerns you have i am afraid, although it could be worth investigating "build system flags vs. manual compile" (atleast i find it interesting).
Again, i am sorry for the noise.
Sveinar
On 29.09.2021 18:18, Sveinar Søpler wrote:
Sveinar
On 29.09.2021 16:42, Eric Pouech wrote:
to correct myself:
- I thought the trigger was "simply" the upgrade to GCC version 11...
after some quick search, Fedora claims moving from 11.0 (F34) to 11.2 (F35), but looks like F34 is pushing gcc 11.2 in updates since July... so the root cause of those new warnings need to be clarified
- the warnings I'm seeing are triggered by -Wmisleading-indentation,
-Warray-bounds, -Wsizeof-array-div, -Wmaybe-uninitialized, which are all active when -Wall is used
it's strange from the log link above that non of the options work with the cross compiler
[ 569s] checking whether the cross-compiler supports -target x86_64-w64-mingw32 -fuse-ld=lld... no [ 569s] checking whether the cross-compiler supports -fno-strict-aliasing... no [ 569s] checking whether the cross-compiler supports -Werror=unknown-warning-option... no [ 569s] checking whether the cross-compiler supports -Werror=ignored-optimization-argument... no [ 569s] checking whether the cross-compiler supports -Wdeclaration-after-statement... no [ 570s] checking whether the cross-compiler supports -Wempty-body... no [ 570s] checking whether the cross-compiler supports -Wignored-qualifiers... no [ 570s] checking whether the cross-compiler supports -Winit-self... no [ 570s] checking whether the cross-compiler supports -Wpacked-not-aligned... no [ 570s] checking whether the cross-compiler supports -Wpragma-pack... no [ 570s] checking whether the cross-compiler supports -Wshift-overflow=2... no [ 570s] checking whether the cross-compiler supports -Wstrict-prototypes... no [ 570s] checking whether the cross-compiler supports -Wtype-limits... no [ 570s] checking whether the cross-compiler supports -Wunused-but-set-parameter... no [ 570s] checking whether the cross-compiler supports -Wvla... no [ 570s] checking whether the cross-compiler supports -Wwrite-strings... no [ 570s] checking whether the cross-compiler supports -Wpointer-arith... no [ 570s] checking whether the cross-compiler supports -Wlogical-op... no [ 570s] checking whether the cross-compiler supports -Wabsolute-value... no [ 570s] checking whether the cross-compiler supports -Wno-format... no [ 570s] checking whether the cross-compiler supports -Wformat-overflow... no [ 570s] checking whether the cross-compiler supports -Wnonnull... no [ 570s] checking whether the cross-compiler supports -mcx16... no [ 570s] checking whether the cross-compiler supports -gdwarf-2... no [ 571s] checking whether the cross-compiler supports -gstrict-dwarf... no [ 571s] checking whether the cross-compiler supports -fexcess-precision=standard... no I'd expect that a couple of them are supported
That was indeed a bit strange. When i do just a regular configure --enable-win64 without any flags, i get this on Ubuntu 21.04:
checking for x86_64-w64-mingw32-gcc... x86_64-w64-mingw32-gcc checking whether x86_64-w64-mingw32-gcc works... yes checking whether the cross-compiler supports -target x86_64-w64-mingw32 -fuse-ld=lld... no checking whether the cross-compiler supports -fno-strict-aliasing... yes checking whether the cross-compiler supports -Werror=unknown-warning-option... no checking whether the cross-compiler supports -Werror=ignored-optimization-argument... no checking whether the cross-compiler supports -Wdeclaration-after-statement... yes checking whether the cross-compiler supports -Wempty-body... yes checking whether the cross-compiler supports -Wignored-qualifiers... yes checking whether the cross-compiler supports -Winit-self... yes checking whether the cross-compiler supports -Wpacked-not-aligned... yes checking whether the cross-compiler supports -Wpragma-pack... no checking whether the cross-compiler supports -Wshift-overflow=2... yes checking whether the cross-compiler supports -Wstrict-prototypes... yes checking whether the cross-compiler supports -Wtype-limits... yes checking whether the cross-compiler supports -Wunused-but-set-parameter... yes checking whether the cross-compiler supports -Wvla... yes checking whether the cross-compiler supports -Wwrite-strings... yes checking whether the cross-compiler supports -Wpointer-arith... yes checking whether the cross-compiler supports -Wlogical-op... yes checking whether the cross-compiler supports -Wabsolute-value... yes checking whether the cross-compiler supports -Wno-format... yes checking whether the cross-compiler supports -Wformat-overflow... yes checking whether the cross-compiler supports -Wnonnull... yes checking whether the cross-compiler supports -mcx16... yes checking whether the cross-compiler supports -gdwarf-2... yes checking whether the cross-compiler supports -gstrict-dwarf... yes checking whether the cross-compiler supports -fexcess-precision=standard... yes
I do kinda fear this pulls us down a unrelated rabbithole - some buildflags or "OBS quirks" perhaps.. dunno. In short: The log from WineHQ buildserver does not really look like it should :)
Sveinar