Musl build failure because of -Wimplicit-function-declaration

[newdev8]

When trying to compile it on a Raspberry Pi 4B with musl I get:

clang  -O3 -pipe -flto -mcpu=cortex-a72+crc -mtune=cortex-a72 -fvectorize -fomit-frame-pointer -fstack-protector-strong -D_FORTIFY_SOURCE=2 -ftrivial-auto-var-init=zero -fsanitize-trap=signed-integer-overflow,unsigned-integer-overflow -fstack-clash-protection -fPIE -fPIC -fno-semantic-interposition -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -I../include -D_GNU_SOURCE -DNO_ANDROID_BACKEND -DUSE_PCRE2 -DPCRE2_CODE_UNIT_WIDTH=8 -DHAVE_STRLCPY -DHAVE_REALLOCARRAY  -c -o sestatus.o sestatus.c
selinux_restorecon.c:441:16: error: variable has incomplete type 'struct stat64'
  441 |         struct stat64 sb;
      |                       ^
selinux_restorecon.c:441:9: note: forward declaration of 'struct stat64'
  441 |         struct stat64 sb;
      |                ^
selinux_restorecon.c:455:10: error: call to undeclared function 'lstat64'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
  455 |                         ret = lstat64(fl->file, &sb);
      |                               ^
selinux_restorecon.c:455:10: note: did you mean 'lstat'?
/usr/include/sys/stat.h:82:5: note: 'lstat' declared here
   82 | int lstat(const char *__restrict, struct stat *__restrict);
      |     ^
clang  -O3 -pipe -flto -mcpu=cortex-a72+crc -mtune=cortex-a72 -fvectorize -fomit-frame-pointer -fstack-protector-strong -D_FORTIFY_SOURCE=2 -ftrivial-auto-var-init=zero -fsanitize-trap=signed-integer-overflow,unsigned-integer-overflow -fstack-clash-protection -fPIE -fPIC -fno-semantic-interposition -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -I../include -D_GNU_SOURCE -DNO_ANDROID_BACKEND -DUSE_PCRE2 -DPCRE2_CODE_UNIT_WIDTH=8 -DHAVE_STRLCPY -DHAVE_REALLOCARRAY  -c -o setenforce.o setenforce.c
2 errors generated.
make[1]: *** [Makefile:178: selinux_restorecon.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/var/tmp/portage/sys-libs/libselinux-3.8.1/work/libselinux-3.8.1-.arm64/src'
make: *** [Makefile:54: all] Error 1

[newdev8]

The following patch fixes it:

--- a/src/selinux_restorecon.c
+++ b/src/selinux_restorecon.c
@@ -26,2 +26,20 @@
 #include <sys/mman.h>
+#ifdef __linux__
+#include <features.h> /* For __GLIBC__ */
+#endif
+
+/*
+ * On non-glibc Linux systems like musl, when _FILE_OFFSET_BITS=64 is defined,
+ * struct stat is already 64-bit, and struct stat64 or lstat64 may not be
+ * provided as distinct entities or may conflict.
+ * Alias them to their standard, 64-bit capable counterparts.
+ * This addresses compilation errors where 'struct stat64' is an incomplete
+ * type and 'lstat64' is an undeclared function.
+ * Ref: Gentoo bug #905711, SELinuxProject/selinux issue #366
+ */
+#if defined(__linux__) &&!defined(__GLIBC__) && \
+    defined(_FILE_OFFSET_BITS) && _FILE_OFFSET_BITS == 64
+#  define stat64 stat
+#  define lstat64 lstat
+#endif
 #include <fts.h>

[zyxhere]

I also hit this in gentoo https://bugs.gentoo.org/956976

The following patch fixes it:

--- a/src/selinux_restorecon.c
+++ b/src/selinux_restorecon.c
@@ -26,2 +26,20 @@
 #include <sys/mman.h>
+#ifdef __linux__
+#include <features.h> /* For __GLIBC__ */
+#endif
+
+/*
+ * On non-glibc Linux systems like musl, when _FILE_OFFSET_BITS=64 is defined,
+ * struct stat is already 64-bit, and struct stat64 or lstat64 may not be
+ * provided as distinct entities or may conflict.
+ * Alias them to their standard, 64-bit capable counterparts.
+ * This addresses compilation errors where 'struct stat64' is an incomplete
+ * type and 'lstat64' is an undeclared function.
+ * Ref: Gentoo bug #905711, SELinuxProject/selinux issue #366
+ */
+#if defined(__linux__) &&!defined(__GLIBC__) && \
+    defined(_FILE_OFFSET_BITS) && _FILE_OFFSET_BITS == 64
+#  define stat64 stat
+#  define lstat64 lstat
+#endif
 #include <fts.h>

Why not post it on the mailing list or open a pull request?

[newdev8]

Being completely honest, I made so many modifications to my system I wan't sure if it was me or if it was actually broken. Thanks for opening the bug and I hope the patch also fixed it for you!

[zyxhere]

On Sunday, 1 June 2025 at 7:19 AM, newdev8 ***@***.***> wrote: newdev8 left a comment (SELinuxProject/selinux#477) Being completely honest, I made so many modifications to my system I wan't sure if it was me or if it was actually broken. Thanks for opening the bug and I hope the patch also fixed it for you!

The patch is good :) But theres also the same bug on libsemanage https://bugs.gentoo.org/938846 Did you not hit that too?

[newdev8]

Indeed. I recently upgraded my system to clang 20 and broke many things while doing so, so I completely forgot about it.

Here's my patch for it:

--- a/src/direct_api.c  2025-03-08 12:00:00
+++ b/src/direct_api.c  2025-03-08 12:01:00
@@ -20,6 +20,7 @@
 #include <stdio.h>
 #include <stdio_ext.h>
 #include <stdlib.h>
 #include <string.h>
+#include <libgen.h>
 #include <unistd.h>
 #include <sys/stat.h>
 #include <sys/types.h>

--- a/src/conf-parse.y  2025-03-08 12:00:00
+++ b/src/conf-parse.y  2025-03-08 12:01:00
@@ -32,6 +32,7 @@
 #include <unistd.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <libgen.h>
 #include <string.h>

Let me know if it also works for you as I had some other ones too which I think might be irrelevant.

[zyxhere]

On Sunday, 1 June 2025 at 4:04 PM, newdev8 ***@***.***> wrote: newdev8 left a comment (SELinuxProject/selinux#477) Indeed. I recently upgraded my system to clang 20 and broke many things while doing so, so I completely forgot about it. Here's my patch for it: --- a/src/direct_api.c 2025-03-08 12:00:00 +++ b/src/direct_api.c 2025-03-08 12:01:00 @@ -20,6 +20,7 @@ #include <stdio.h> #include <stdio_ext.h> #include <stdlib.h> #include <string.h> +#include <libgen.h> #include <unistd.h> #include <sys/stat.h> #include <sys/types.h> --- a/src/conf-parse.y 2025-03-08 12:00:00 +++ b/src/conf-parse.y 2025-03-08 12:01:00 @@ -32,6 +32,7 @@ #include <unistd.h> #include <stdio.h> #include <stdlib.h> +#include <libgen.h> #include <string.h> Let me know if it also works for you as I had some other ones too which I think might be irrelevant.

The patch works for me with gcc 15

[zyxhere]

Sent with Proton Mail secure email.

On Sunday, 1 June 2025 at 3:11 PM, ***@***.*** ***@***.***> wrote: On Sunday, 1 June 2025 at 7:19 AM, newdev8 ***@***.*** wrote: > newdev8 left a comment (SELinuxProject/selinux#477) > > Being completely honest, I made so many modifications to my system I wan't sure if it was me or if it was actually broken. Thanks for opening the bug and I hope the patch also fixed it for you! The patch is good :) But theres also the same bug on libsemanage https://bugs.gentoo.org/938846 Did you not hit that too?

Could you change the bug title to indicate that this is because of -Wimplicit-function-declaration on musl regardless of architecture

[peppergrayxyz]

There is another issue after patching stat64 and lstat64 in selinux_restorecon.c (which can also be achived by passing -D_LARGEFILE64_SOURCE see https://musl.libc.org/releases.html)

ld.lld: error: version script assignment of 'LIBSELINUX_3.8' to symbol 'matchpathcon_filespec_add64' failed: symbol not defined
clang: error: linker command failed with exit code 1 (use -v to see invocation)

which is caused by

selinux/libselinux/src/libselinux.map

Line 258 in a5763ea

matchpathcon_filespec_add64;

I fixed this now by removing the 64 suffix in both files (47d5c72 8d40374), which allows me to emerge sys-libs/libselinux-3.8.1 on gentoo (musl/llvm).

What is the proper way to fix this? I have found multiple variants in other repos.

[peppergrayxyz]

The problem with basename in libsemanage has been patched, but this issue wasn't tagged: a339594

[zyxhere]

Like I said this needs to be posted on the SELinux mailing list. They don't use github for anything other than pull requests