[libc-commits] [PATCH] D84611: [libc] Adds fuzz test for strstr and alphabetizes string fuzz CMakeList.

[Chris Gyurgyik via Phabricator via libc-commits]

cgyurgyik added inline comments.


================
Comment at: libc/fuzzing/string/strstr_fuzz.cpp:22
+// its used as the verification case.
+char *BruteForceStrStr(const char *haystack, const char *needle) {
+  const size_t haystack_size = __llvm_libc::strlen(haystack);
----------------
cgyurgyik wrote:
> cgyurgyik wrote:
> > sivachandra wrote:
> > > I am not sure if we should use another implementation of `strstr` to test the real implementation. Instead, can we use `strcmp` in a brute force manner as `strcmp` has its own fuzz test.
> > > 
> > > ```
> > > auto result = __llvm_libc::strstr(haystack, needle);
> > > // Because of the construction below, the input for which result is NULL can be tested separately.
> > > // For the rest:
> > > if (strcmp(result, needle) != 0) {
> > >   __builtin_trap():
> > > }
> > > auto *haystack_ptr = haystack;
> > > for (;haystack_ptr != result; ++haystack)
> > >     if (strcmp(haystack_ptr, needle) == 0) {
> > >       // An earlier occurance of needle was missed by strstr.
> > >       __builtin_trap();
> > >     }
> > > }
> > > ```
> > > 
> > > This is brute force and calls `strcmp` for every byte, but avoids reimplementing `strstr`. WDYT?
> > PTAL and let me know if this is what you're looking for.
> > 
> > I had to make some slight modifications since the `result` returned will include the rest of the string, and thus strcmp() will not always work.
> Hmm nevermind I need to make some changes to this.
As discussed, `strcmp` won't quite work here. Created a helper function that mimics `memcmp` by comparing each letter up to N.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D84611/new/

https://reviews.llvm.org/D84611