Fixes related to type and function specialization #3458
Conversation
|
@fruffy It looks like gauntlet does not like one of the new tests. |
|
Gauntlet complains about this transformation: bool g_0(in h_0 a) {
@name("v") h<h_0> v_0;
v_0.f = a;
return v_0.f == a;
}
bool gg() {
bool tmp;
@name("a") h_0 a_0;
a_0 = (h_0){f = 1w0};
tmp = g_0(a_0);
return tmp;
}
control c(out bool x) {
apply {
x = gg();
}
}to bool g_0(in h_0 a) {
@name("v") h<h_0> v_0;
;
return v_0.f == a;
}
bool gg() {
bool tmp;
@name("a") h_0 a_0;
a_0 = (h_0){f = 1w0};
tmp = g_0(a_0);
return tmp;
}
control c(out bool x) {
apply {
x = gg();
}
}Removing the |
|
Gauntlet seems right, this looks like a bug, but it's probably somewhere else. |
137754b to
08961aa
Compare
|
@fruffy indeed, Gauntlet was right. I have fixed that bug, and now this is ready for review. |
backends/ebpf/targets/xdp_target.py
Outdated
| @@ -0,0 +1 @@ | |||
| /home/mbudiu/git/p4c/extensions/p4c-xdp/xdp_target.py | |||
There was a problem hiding this comment.
There must be a way to ignore symlinks in specific folders. Let me check.
Signed-off-by: Mihai Budiu <mbudiu@vmware.com>
Signed-off-by: Mihai Budiu <mbudiu@vmware.com>
Signed-off-by: Mihai Budiu <mbudiu@vmware.com>
Signed-off-by: Mihai Budiu <mbudiu@vmware.com>
Signed-off-by: Mihai Budiu <mbudiu@vmware.com>
62a7660 to
5e7c5d2
Compare
Signed-off-by: Mihai Budiu <mbudiu@vmware.com>
Signed-off-by: Mihai Budiu mbudiu@vmware.com
Fixes #3291
The fix is in the typechecking code.
But the tests revealed some bugs in the specialization code: now we insert the specialized function code in a different place in the source code, to make sure it's after all type declarations that it may depend on. So the function specialization code is more similar to the existing type specialization code.