[PATCH] libsanitizer: Support non-POSIX extension REG_STARTEND.

[marxin]

Summary

A recent (15 years old) glibc is needed in order to support
REG_STARTEND.

The patch addresses problem reported here:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98920

[vitalybuka]

There is D96348

[marxin]

In D95864#2573793, @vitalybuka wrote:

There is D96348

Ahh, good. That's addressing the very same issue.

[arichardson]

compiler-rt/lib/sanitizer_common/sanitizer_platform.h

341

Is this missing SANITIZER_I386?

Having to hardcode all the architectures is rather awkward, maybe we should just fall back to an unversioned lookup if the versioned one fails?

[marxin]

compiler-rt/lib/sanitizer_common/sanitizer_platform.h

341

Yes, doing an unversioned fallback would be the best approach! Can you please do it in your patch submission?
I tend to close this review request.

compiler-rt/lib/sanitizer_common/sanitizer_common_interceptors.inc

 }
 INTERCEPTOR(void, regfree, const void *preg) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, regfree, preg);
   if (preg)
     COMMON_INTERCEPTOR_READ_RANGE(ctx, preg, struct_regex_sz);
   REAL(regfree)(preg);
 }
+
+#ifdef SANITIZER_REGEXEC_VERSION
+#define COMMON_INTERCEPT_FUNCTION_REGEXEC(fn)                          \
+    COMMON_INTERCEPT_FUNCTION_VER(fn, SANITIZER_REGEXEC_VERSION)
+#else
+#define COMMON_INTERCEPT_FUNCTION_REGEXEC(fn)                          \
+    COMMON_INTERCEPT_FUNCTION(fn)
+#endif
+
 #define INIT_REGEX                                                             \
   COMMON_INTERCEPT_FUNCTION(regcomp);                                          \
-  COMMON_INTERCEPT_FUNCTION(regexec);                                          \
+  COMMON_INTERCEPT_FUNCTION_REGEXEC(regexec);                                  \
   COMMON_INTERCEPT_FUNCTION(regerror);                                         \
   COMMON_INTERCEPT_FUNCTION(regfree);
 #else
 #define INIT_REGEX
 #endif
 
 #if SANITIZER_INTERCEPT_REGEXSUB
 INTERCEPTOR(SSIZE_T, regnsub, char *buf, SIZE_T bufsiz, const char *sub,

compiler-rt/lib/sanitizer_common/sanitizer_platform.h

 // On linux, some architectures had an ABI transition from 64-bit long double
 // (ie. same as double) to 128-bit long double.  On those, glibc symbols
 // involving long doubles come in two versions, and we need to pass the
 // correct one to dlvsym when intercepting them.
 #if SANITIZER_LINUX && (SANITIZER_S390 || SANITIZER_PPC32 || SANITIZER_PPC64V1)
 #define SANITIZER_NLDBL_VERSION "GLIBC_2.4"
 #endif
 
+// On linux, we need a recent regexec implementation that supports
+// a non-POSIX flag REG_STARTEND.
+#if SANITIZER_LINUX
+#define SANITIZER_REGEXEC_VERSION "GLIBC_2.3.4"
+#endif
+
 #if SANITIZER_GO == 0
 # define SANITIZER_GO 0
 #endif
 
 // On PowerPC and ARM Thumb, calling pthread_exit() causes LSan to detect leaks.
 // pthread_exit() performs unwinding that leads to dlopen'ing libgcc_s.so.
 // dlopen mallocs "libgcc_s.so" string which confuses LSan, it fails to realize
 // that this allocation happens in dynamic linker and should be ignored.