Fixing unit tests for mem-cpy, -move and -set functions #112
Conversation
* mod: Disable memcpy interposition by default Signed-off-by: Diogo Behrens <[email protected]> * cicd: Trigger pipeline on pull requests Signed-off-by: Diogo Behrens <[email protected]> * cicd: Enable memcpy interposition where safe Signed-off-by: Diogo Behrens <[email protected]> --------- Signed-off-by: Diogo Behrens <[email protected]>
Signed-off-by: Diogo Behrens <[email protected]>
Signed-off-by: gigna <[email protected]>
Signed-off-by: gigna <[email protected]>
…ng. enable function did not replace the actual call to functions. For this reason I used #define mem_func fake_mem_func to tell the preprocessor to replace every call to mem func with calls to the fake functions.
|
This is not passing in alpine |
…the intended return values of the mocks.
Signed-off-by: gigna <[email protected]>
Signed-off-by: georgeta-igna <[email protected]>
test/interpose/memcpy_test.c
Outdated
| struct memcpy_event *ev = EVENT_PAYLOAD(ev); | ||
| log_printf("ev->dest %p\n", &ev->dest); | ||
| log_printf("&(&E_memcpy)->dest %p\n", &(&E_memcpy)->dest); | ||
| log_printf("memcmp(&ev->dest, &(&E_memcpy)->dest, sizeof((&E_memcpy)->dest)) == 0 %d\n", memcmp(&ev->dest, &(&E_memcpy)->dest, sizeof(__typeof((&E_memcpy)->dest))) == 0); |
There was a problem hiding this comment.
should we really commit these log messages?
If yes, I recommend defining a macro to avoid writing the same text twice
There was a problem hiding this comment.
no, I am not planning to merge with these. I want to understand better why for the arm with clang the next ensure fails: https://github.com/open-s4c/dice/actions/runs/18745327427/job/53480906689
| ASSERT_FIELD_EQ(&E_memset, ptr); | ||
| ASSERT_FIELD_EQ(&E_memset, value); | ||
| ASSERT_FIELD_EQ(&E_memset, num); | ||
| ASSERT_FIELD_EQ(&E_memset, ret); |
| E_memset.ptr = malloc(5); | ||
| E_memset.value= 3; | ||
| E_memset.num = 2; | ||
| E_memset.ret = E_memset.ptr; |
There was a problem hiding this comment.
I think they are always equal, so we should probably remove this .ret field and just compare with E_memset.ptr
There was a problem hiding this comment.
E_memset.ret is null without this and it is checked somewhere in the main test: ensure(ret == E_memset.ptr);
There was a problem hiding this comment.
I mean there's no point to have the field at all, it can be removed from the struct. Then you can still ensure(ret == E_memset.ptr);
test/interpose/memcpy_test.c
Outdated
| enable(fake_memset); | ||
| E_memset.ptr = malloc(5); | ||
| E_memset.value= 3; | ||
| E_memset.num = 2; |
There was a problem hiding this comment.
maybe better to use compound literal for this initialization?
| log_printf("&(&E_memcpy)->dest %p\n", &(&E_memcpy)->dest); | ||
| log_printf("memcmp(&ev->dest, &(&E_memcpy)->dest, sizeof((&E_memcpy)->dest)) == 0 %d\n", memcmp(&ev->dest, &(&E_memcpy)->dest, sizeof(__typeof((&E_memcpy)->dest))) == 0); | ||
| ASSERT_FIELD_EQ(&E_memcpy, dest); | ||
| ASSERT_FIELD_EQ(&E_memcpy, src); |
There was a problem hiding this comment.
can we add some check that the interceptor really got called? Or is that already checked elsewhere?
…eturn value of these functions as for the original memory handling functions
… from interpose/REAL
Signed-off-by: georgeta-igna <[email protected]>
|
Replaced by #122 |
4 participants
No description provided.