On Sun Aug 11, 2024 at 4:56 PM CEST, Reinhard Tartler wrote: > On 2024-08-11 08:55, Diederik de Haas wrote: > > On 03 Aug 2024 11:08:58 -0400 Reinhard Tartler <siret...@tauware.de> > > wrote: > >> I noticed this package is listed as low-NMU. As such, I'm taking the > >> liberty of uploading the following patch as NMU to sid: > >> ... > >> new file debian/patches/don-t-fail-on-unknown-gcc-warnings.patch > >> @@ -0,0 +1,20 @@ > >> +From: Reinhard Tartler <siret...@tauware.de> > >> +Date: Sat, 3 Aug 2024 10:46:50 -0400 > >> +Subject: don't fail on unknown gcc warnings > >> + > >> +--- > >> + cmake/Helper.cmake | 1 - > >> + 1 file changed, 1 deletion(-) > >> + > >> +diff --git a/cmake/Helper.cmake b/cmake/Helper.cmake > >> +index f9cdcf2..126e93f 100644 > >> +--- a/cmake/Helper.cmake > >> ++++ b/cmake/Helper.cmake > >> +@@ -39,7 +39,6 @@ else() > >> + > >> + if(NOT WASMEDGE_PLUGIN_WASI_NN_GGML_LLAMA_CUBLAS) > >> + list(APPEND WASMEDGE_CFLAGS > >> +- -Werror > >> + -Wno-error=pedantic > >> + ) > >> + if(CMAKE_CXX_COMPILER_ID MATCHES "GNU" AND > >> CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 13) > >> new file debian/patches/series > >> @@ -0,0 +1 @@ > >> +don-t-fail-on-unknown-gcc-warnings.patch > > > > Why do you consider this an appropriate solution? > > > > Upstream explicitly want all warnings to be treated as errors and now > > with gcc-14 it generates a new warning. > > This sounds like something upstream explicitly wants to know about? > > I think this is a reasonable stance to take upstream. I've now filed > https://github.com/WasmEdge/WasmEdge/issues/3640 to document this issue, > in the hope that someone with more expertise can have a look.
Thanks for reporting it upstream :-) > For Debian, I do think that this workaround is acceptable, at least for > the purposes of allowing further testing in the "testing" Distribution, > so that we get additional datapoints whether there actually are runtime > issues that stem from unitialized variables that GCC claims to detect. I disagree, but I'll let it up to the maintainer to reopen this bug or not. Packages don't transition to Testing to get additional datapoints (even though that will happen), but because there are no RC bugs, like FTBFS, currently known. That's not the case here. If you can make the argument that a specific warning can be ignored, you could override that specific warning with a clear explanation why that's OK in this particular case. This patch OTOH essentially says to ignore ALL warnings. My 0.02
signature.asc
Description: PGP signature
