fs: Add missing retarget definitions for ARM/IAR #5204
Conversation
geky
force-pushed
the
fix-retarget-defs
branch
from
81ba72d to
7ea9811
Compare
|
cc @kjbracey-arm @deepikabhavnani |
There was a problem hiding this comment.
No objections, with only the passing note that the POSIX spec would like O_RDONLY to be a proper bit, despite historical implementations having it 0. I guess we want to match glibc though.
Where did the "exhaustive" errno list come from? POSIX or glibc or what?
|
/morph test |
|
@geky Does this need anything more than a test? |
|
/morph test-nightly |
1 similar comment
|
/morph test-nightly |
Result: FAILUREYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
Result: FAILUREYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
Result: FAILUREYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
Result: FAILUREYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
Result: FAILUREYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
Result: FAILUREYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
Result: FAILUREYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
|
@geky Please look at the failures ( |
theotherjimmy
changed the title
- missing O_ flags - missing S_IF flags - missing system types - missing stat structure - exhaustive list of errno codes
geky
force-pushed
the
fix-retarget-defs
branch
from
7ea9811 to
e9cd239
Compare
|
@0xc0170,Thanks for the line number
Ah, that's why ssize_t was defined to int. The LocalFileSystem incorrectly uses int in its overload, as opposed to the ssize_t typedef. But if this breaks something in our codebase, it's probably safer to just keep the old typedef. updated |
Result: ABORTEDYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
|
/morph test-nightly |
Result: FAILUREYour command has finished executing! Here's what you wrote!
OutputBuild failed! |
|
/morph test-nightly |
|
@kjbracey-arm, sorry I missed your message under all the morph test outputs I just stole the errno codes from lxr, though they're spread between a few files: The O_RDONLY thing is noted, though yeah we're just aligning with glibc/newlib |
Result: SUCCESSYour command has finished executing! Here's what you wrote!
OutputAll builds and test passed! |
Currently code written for GCC may not work on ARM/IAR due to missing retarget definitions. This is manifested in littlefs/spiffs and may impact users.
These were missing and causing tests written against the filesystem to fail. Tests are inbound for a later pr. Currently these definitions are available in GCC, but missing in ARM/IAR. This patch fixes that.
Note: Yes there are a bunch of errno codes that we may never use, but that just means we won't have to worry about naming errors for a while now.