pkg/csource: use defines from sys/syscall.h on *bsd, fixes #841

[blackgnezdo]

No description provided.

[blackgnezdo]

I checked that csource_test passes on openbsd. @tuexen if you could help with testing on freebsd, it would be most appreciated.

[tuexen]

I checked that csource_test passes on openbsd. @tuexen if you could help with testing on freebsd, it would be most appreciated.

@blackgnezdo I'm more than happy to test this tomorrow.

[tuexen]

@blackgnezdo This PR compiles fine on FreeBSD (gmake extract / gmake generate / gmake). Should I test anything else?

[dvyukov]

go test -short ./pkg/csource will test this, but I am not sure we ever tested it on freebsd, but it may not work for other reasons

[tuexen]

It works with the following patch:

diff --git a/sys/targets/targets.go b/sys/targets/targets.go
index 88d2c727..43d2c513 100644
--- a/sys/targets/targets.go
+++ b/sys/targets/targets.go
@@ -274,7 +274,7 @@ var oses = map[string]osCommon{
                ExecutorUsesShmem:      true,
                ExecutorUsesForkServer: true,
                KernelObject:           "vmlinux",
-               CPP:                    "cpp",
+               CPP:                    "g++",
        },
        "netbsd": {
                SyscallNumbers:         true,

since on FreeBSD cpp maps to clang, which does not support the -fdirectives-only.

With the above change the following happens:

tuexen@epyc:~/go/src/github.com/google/syzkaller % go test -short ./pkg/csource
ok      github.com/google/syzkaller/pkg/csource 5.725s

@dvyukov Is the fix correct? If that is the case, can you commit the fix or should I provide a PR?

[dvyukov]

Is the fix correct? If that is the case, can you commit the fix or should I provide a PR?

Done:
31677db
Thanks