xtables-addons build failed

[ptpt52]

since openwrt/openwrt#12545 merged

  CC [M]  /mnt/Data/Sources/openwrt/x-wrt/build_dir/target-aarch64_cortex-a53_musl/linux-mediatek_filogic/xtables-addons-3.21/extensions/LUA/lua/lcode.o
  CC [M]  /mnt/Data/Sources/openwrt/x-wrt/build_dir/target-aarch64_cortex-a53_musl/linux-mediatek_filogic/xtables-addons-3.21/extensions/LUA/lua/ldebug.o
In file included from ./include/linux/string.h:9,
                 from /mnt/Data/Sources/openwrt/x-wrt/build_dir/target-aarch64_cortex-a53_musl/linux-mediatek_filogic/xtables-addons-3.21/extensions/LUA/lua/include/string.h:1,
                 from /mnt/Data/Sources/openwrt/x-wrt/build_dir/target-aarch64_cortex-a53_musl/linux-mediatek_filogic/xtables-addons-3.21/extensions/LUA/lua/ldebug.c:13:
./include/linux/stdarg.h:6: error: "va_start" redefined [-Werror]
    6 | #define va_start(v, l)  __builtin_va_start(v, l)
      | 
In file included from /mnt/Data/Sources/openwrt/x-wrt/build_dir/target-aarch64_cortex-a53_musl/linux-mediatek_filogic/xtables-addons-3.21/extensions/LUA/lua/ldebug.c:8:
/mnt/Data/Sources/openwrt/x-wrt/staging_dir/toolchain-aarch64_cortex-a53_gcc-12.2.0_musl/lib/gcc/aarch64-openwrt-linux-musl/12.2.0/include/stdarg.h:47: note: this is the location of the previous definition
   47 | #define va_start(v,l)   __builtin_va_start(v,l)
      | 
./include/linux/stdarg.h:8: error: "va_arg" redefined [-Werror]
    8 | #define va_arg(v, T)    __builtin_va_arg(v, T)
      | 
/mnt/Data/Sources/openwrt/x-wrt/staging_dir/toolchain-aarch64_cortex-a53_gcc-12.2.0_musl/lib/gcc/aarch64-openwrt-linux-musl/12.2.0/include/stdarg.h:49: note: this is the location of the previous definition
   49 | #define va_arg(v,l)     __builtin_va_arg(v,l)
      | 
./include/linux/stdarg.h:9: error: "va_copy" redefined [-Werror]
    9 | #define va_copy(d, s)   __builtin_va_copy(d, s)
      | 
/mnt/Data/Sources/openwrt/x-wrt/staging_dir/toolchain-aarch64_cortex-a53_gcc-12.2.0_musl/lib/gcc/aarch64-openwrt-linux-musl/12.2.0/include/stdarg.h:52: note: this is the location of the previous definition
   52 | #define va_copy(d,s)    __builtin_va_copy(d,s)
      | 
cc1: all warnings being treated as errors
scripts/Makefile.build:289: recipe for target '/mnt/Data/Sources/openwrt/x-wrt/build_dir/target-aarch64_cortex-a53_musl/linux-mediatek_filogic/xtables-addons-3.21/extensions/LUA/lua/ldebug.o' failed

[ptpt52]

ping @AuthorReflex @neheb
any help to make this build failure?

[Ansuel]

@ptpt52 i found i very similar problem...

it's either we need to fix your include in toolchain to match (so the redefine gets automagically muted since nothing is redefined)

or add and ifdef...

in practice toolchain include for stdarg is mixing with the kernel one resulting in this redefine error.

@mpratt14 might help in sorting this mess..

Current fix i used was https://github.com/openwrt/openwrt/blob/3ccdc15c6ab14df898d9f36f0d7fa752ac1c15d6/package/kernel/broadcom-wl/patches/919-fix-compilation-warning-for-stack-limit.patch#L294

[ptpt52]

@Ansuel
thanks for tips
I try out a patch(https://github.com/x-wrt/packages/blob/master/net/xtables-addons/patches/300-fix-variable-length-array.patch) to fix the compilation

[Ansuel]

does it work?

[ptpt52]

does it work?

it compile success for my case. but the changes looks not that elegant

[Ansuel]

@ptpt52 i would split the 2 patch... for the define we can't do that much... (i think the include should just be exploded from the include dir... @mpratt14 might help)

for the variable array thing... they are kernel modules right?

i would use kzalloc and also it should be checked for error checking if the alloc failed and returning -ENOMEM

(I did plenty of these change in the WERROR pr so you should have a direct example of the pattern to use for such change. If not feel free to ask)

Variable array change is clean if done with this pattern...
for the define thing only way saddly (if the include can't be removed)

[Ansuel]

Any news for a modified patch?

[ptpt52]

@Ansuel could you help to create a new one since I have no time for now.

[pprindeville]

There's no handling of kmalloc() returning NULL.

[Ansuel]

@pprindeville already pointed out, i'm preparing a patch to address my comments

[pprindeville]

BTW, could we (re)generate Mbuild dynamically based on what modules are selected?

If one of the modules is unhealthy, it would be nice to be able to deselect it and have a good build...