Re: [libvirt PATCH 5/5] tests: Allow expansion of mocked stat symbols

On Mon, Mar 07, 2022 at 10:01:40AM +0000, Daniel P. Berrangé wrote: > On Mon, Mar 07, 2022 at 10:04:06AM +0100, Martin Kletzander wrote: > > When libc uses a define to rewrite stat64 to stat our mocks do not work if they > > are chained because the symbol that we are looking up is being stringified and > > therefore preventing the stat64->stat expansion per C-preprocessor rules. One > > stringification macro is just enough to make it work. > > This doesn't sound right to me. > > If we're implementing a mock for 'stat64', we should not be looking > up 'stat'. We should be implmenting a mock for 'stat' instead. > > It sounds more like we got MOCK_STAT and MOCK_STAT64 incorrectly > defined surely. > I spent three days trying to figure this out. But with this fix we are actually doing what you are suggesting. All the (l)stat64 are being replaced by (l)stat, so even: int lstat64(const char *path, struct stat64 *statbuf) is actually rewritten to int lstat(const char *path, struct stat *statbuf) because the define applies to the struct name as well. The only difference is that internally it will use the real_lstat64 function pointer, but the function declaration is correct. And if you enable debugging it will say that lstat64 is being redirected, but that's a pretty minor issue, if it is one at all. What we could do is rewrite the macros so that all of the above is properly expanded/substituted, but I do not think that is the way we want to do things as it looked a bit messy (I probably still have that commit somewhere.