Windows compilation error #85774
Comments
tejlmand
added
priority: medium
|
@ofirshe thanks for the CMake trace. Both from the original issue error log and your trace I have traced down the issue to be related to the generator expressions used by the llext subsys here: Line 2239 in ee37a94 The genex expression is part of a and it seems to be the expansion of the genex in However, I cannot tell from the trace itself what the genex expanded into. I have made a small a debug patch file to help further investigate this issue. diff --git a/CMakeLists.txt b/CMakeLists.txt
index 16a471ae6ee..93701dea7bc 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -2238,6 +2238,9 @@ list(APPEND llext_edk_cflags ${LLEXT_EDK_APPEND_FLAGS})
build_info(llext-edk file PATH ${llext_edk_file})
build_info(llext-edk cflags VALUE ${llext_edk_cflags})
build_info(llext-edk include-dirs VALUE "$<TARGET_PROPERTY:zephyr_interface,INTERFACE_INCLUDE_DIRECTORIES>")
+message("llext-cflags genex : ${llext_edk_cflags}")
+add_custom_target(print_llext_cflags COMMAND ${CMAKE_COMMAND} -E echo \"${llext_edk_cflags}\")
+add_custom_target(print_llext_includes COMMAND ${CMAKE_COMMAND} -E echo \"$<TARGET_PROPERTY:zephyr_interface,INTERFACE_INCLUDE_DIRECTORIES>\")
add_custom_command(
OUTPUT ${llext_edk_file}This patch provides to custom targets which helps us understand exactly how the genex are expanded on your system. To get the expanded values, then run: and provide the output in this issue. You can also do it using FYI @pillo79 seems we have some issues on Windows with expanding genex'es in certain cases. |
|
@tejlmand the JSON files should definitely not have genexes in it, they are written with |
@pillo79 Right, it was not the file written to disk, but the input that was going into |
Which I guess will just downgrade this issue from medium to minor 😛 |
Indeed. 😢
Do you think any of these is workable? Or maybe there's some magic |
|
Hey @tejlmand, Regarding your request, could you clarify the exact steps you'd like me to run? I'm working with Zephyr as a submodule in an out-of-tree setup:
Could you please provide precise instructions? I want to ensure I follow your request correctly. |
Let me try.
and provide the output in this issue 🙂 |
|
Build output: print_llext_cflags output: print_llext_includes output: @tejlmand On the latest Zephyr main branch, this issue is no longer reproducible (I assume recent changes have resolved it). |
|
@ofirshe thanks for this output, very helpful!
Yes, the PR I mentioned above was just merged and avoids the bug unless you specifically opt in for the EDK functionality. So the bug is still there, but not visible to most users. |
|
@pillo79 I include the entire |
|
Thank you... that effectively rules out the "easy fix" I had prepared (#85884). 😢 |
kartben
added
bug
|
Discussed in release WG meeting and rising priority to release blocker as this sounds like this impacts multiple users. |
@ofirshe thanks for the details. and it seems that it's your own code which has expanded into I wonder how exactly you're prepending the current path, because Can you provide an exact copy of line 5 and 6 from |
|
Hey @tejlmand, I've developed several custom drivers that I use as Zephyr modules, and you're right. Inside those modules, I include header files using |
@ofirshe There's nothing wrong in using The important part right now is to understand from exactly where the backslahes originates, which is why I would still like you to post the two CMake lines in question.
|
where |
Discussed in #85747
Originally posted by ofirshe February 13, 2025
Hi, I've rebased onto the latest Zephyr commit (8b80c20) and I'm encountering compilation errors related to YAML and JSON escaping that I've never seen before.
Can someone help me debug this?
The text was updated successfully, but these errors were encountered: