On Thu, Jan 28, 2021 at 11:24:30AM +0100, Tim Wiederhake wrote:
clang-tidy is a static code analysis tool under the llvm umbrella. It
is
primarily meant to be used on C++ code bases, but some of the checks it
provides also apply to C.
The findings vary in severity and contain pseudo-false-positives, i.e.
clang-tidy is flagging potential execution flows that could happen in
theory but are virtually impossible in real life: In function
`virGetUnprivSGIOSysfsPath`, variables `maj` and `min` would be read
unintialized if `stat()` failed and set `errno` to a negative value, to name
just one example.
The main source of false positive findings is the lack of support for
`__attribute__((cleanup))` in clang-tidy, which is heavily used in libvirt
through glib's `g_autofree` and `g_auto()` macros:
#include <stdlib.h>
void freeptr(int** p) {
if (*p)
free(*p);
}
int main() {
__attribute__((cleanup(freeptr))) int *ptr = NULL;
ptr = calloc(sizeof(int), 1);
return 0; /* flagged as memory leak of `ptr` */
}
This sadly renders clang-tidy's analysis of dynamic memory useless, hiding all
real issues that it could otherwise find.
Meson provides excellent integration for clang-tidy (a "clang-tidy" target is
automatically generated if a ".clang-tidy" configuration file is present
in the project's root directory). The amount of false-positives and the slow
analysis, triggering time-outs in the CI, make this tool unfit for inclusion
in libvirt's GitLab CI though.
Is it possible to make it viable for CI by disabling *all* checks by
default and then selectively re-enabling just the handful that are
useful and don't have false positives ?
Regards,
Daniel
--
|:
https://berrange.com -o-
https://www.flickr.com/photos/dberrange :|
|:
https://libvirt.org -o-
https://fstop138.berrange.com :|
|:
https://entangle-photo.org -o-
https://www.instagram.com/dberrange :|