-
Notifications
You must be signed in to change notification settings - Fork 75
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fix CMake compiler option handling #989
fix CMake compiler option handling #989
Conversation
On windows it complains: |
Where do you get this error from? I do not see any broken windows CI test. Is it on your machine? |
No, it is the 7th build from the bottom (github action) |
Currenlty we split very often the compiler option and the argument. This creates issues when the compiler parses the given arguments. The fix is that the option and argument is explicit grouped.
8e9bbf5
to
08ad127
Compare
I will check it maybe it is not allowed to embed the ptx code when seperateable compile is selected.
Am 27. April 2020 19:28:39 MESZ schrieb Benjamin Worpitz <notifications@github.com>:
…No, it is the 7th build from the bottom (github action)
--
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub:
#989 (comment)
|
The problem is when separate able compile is activated CMake is keeping the quotes
|
When |
I will close this issue. The problem was triggered by our third party project in PIConGPU. The problem why Normally removing duplicated options should not break the build system. The problem is that we add options normally wrong. I tried to solve it in this PR but when enabling |
This is the replacement PR for alpaka-group#989 Currenlty we split very often the compiler option and the argument. This creates issues when the compiler parses the given arguments. The fix is that the compiler option and argument will be passed with `=` instead of a space in between.
This is the replacement PR for alpaka-group#989 Currenlty we split very often the compiler option and the argument. This creates issues when the compiler parses the given arguments. The fix is that the compiler option and argument will be passed with `=` instead of a space in between.
This is the replacement PR for alpaka-group#989 Currenlty we split very often the compiler option and the argument. This creates issues when the compiler parses the given arguments. The fix is that the compiler option and argument will be passed with `=` instead of a space in between.
This is the replacement PR for #989 Currenlty we split very often the compiler option and the argument. This creates issues when the compiler parses the given arguments. The fix is that the compiler option and argument will be passed with `=` instead of a space in between.
Currenlty we split very often the compiler option and the argument.
This creates issues when the compiler parses the given arguments.
CMake or we in alpaka remove duplicated options and creates therefore wrong argument lines:
example:
-Xcudafe --display_error_number -Xcudafe --diag_suppress=esa_on_defaulted_function_ignored
was changed to
-Xcudafe --display_error_number --diag_suppress=esa_on_defaulted_function_ignored
and results in an error:
nvcc fatal : Unknown option '-diag_suppress'
The fix is that the option and argument is explicit grouped.
This bug is included in 0.4.0 too, a backport must be manually applied because we changed our CMake structure in the develop branch.
I found the bug during tests with docker for PIConGPU.
If the bug becomes visible depends on the order of arguments. In the worst case an option could be ignored by the compiler without throwing an error.