Enforce Werror when building samples on Linux #3816
Conversation
|
bors try |
tryBuild failed: |
| -fno-builtin>) | ||
| -fno-builtin> | ||
| # Treat warnings as errors by default. Application based on OE SDK | ||
| # will inhere this flag, which is consistent to the behavior of the |
There was a problem hiding this comment.
| # will inhere this flag, which is consistent to the behavior of the | |
| # will inherit this flag, which is consistent to the behavior of the |
| @@ -275,7 +275,11 @@ if (OE_SGX) | |||
| # user. There are valid reasons for an end user to use built-ins. | |||
| $<BUILD_INTERFACE:-fno-builtin-malloc | |||
| -fno-builtin-calloc | |||
| -fno-builtin>) | |||
| -fno-builtin> | |||
| # Treat warnings as errors by default. Application based on OE SDK | |||
There was a problem hiding this comment.
Is this desirable? This means that every application that uses liboecore via CMake import or pkgconfig will get -Werror by default, not just the samples. It's not generally my experience that SDKs set this on behalf of a dev (folks should feel to correct this if I am mistaken.)
This is also potentially a public build breaking change and should at least be documented in the CHANGELOG.md.
There was a problem hiding this comment.
Good point. I'm unapproving till we are sure that we don't force these flags on the users.
There was a problem hiding this comment.
There is a distinction:
- We want the samples to be built using -Werror
- However, we don't want to force -Werror on the user.
My understanding is that, today, we don't force the flags on the user on both Linux and Windows.
There was a problem hiding this comment.
No. -Werror is added in the clangw by default, so samples on Windows are always compiled with the flag, which is what the #3773 about.
There was a problem hiding this comment.
Thanks for the clarification. We were forcing the flag on Windows, whereas we weren't on Linux.
I'm removing my blocker.
It is a matter of consistency then.
There was a problem hiding this comment.
Exactly. Given that @CodeMonkeyLeet pointed out that we shouldn't enforce -Werror on behalf of users, I'm wondering how should we address the consistency? This PR currently follows the conclusion from triaging #3733. We could also remove the enforcement on the Windows side. Thoughts?
There was a problem hiding this comment.
There are convincing arguments for both choices. Thinking about it bit more, my preference would be to enforce -Werror since it is generally considered a good programming practice to compile code with high warning levels.
There was a problem hiding this comment.
+1 to Anand's summary before about what is desired state:
- We want the samples to be built using -Werror
- However, we don't want to force -Werror on the user.
Above should be true for both Linux and Windows.
dthaler
added
devex
Signed-off-by: Ming-Wei Shih <mishih@microsoft.com>
dthaler
force-pushed
the
enforce_werror_to_linux_sampes
branch
from
e5424c4
to
704c93c
Compare
|
@dthaler, please review |
|
bors try |
tryBuild failed: |
This PR fixes #3773, which enforces
Werrorflag when building samples on Linux.Signed-off-by: Ming-Wei Shih mishih@microsoft.com