Build.md gives inaccurate build instructions on POSIX and should be fixed #111
Labels
Comments
|
|
Yes, this part was highlighted, but a) I don't know where that problem lies / how to fix it (if it's in that project or in the loading of that project in llfio), and b) I didn't find any build information for QuickCppLib on its own, so I couldn't replicate that. If you can give me any further information/links to that, I will try to replicate it, and I can also open the issue on the other project if that's necessary for you, I just wasn't sure because I didn't compile that on its own specifically. |
|
cmake needs to be told what kind of build you want. For historical reasons, it's still possible to choose "none" and that leads to builds which cause people to report issues here which are the fault of cmake, not me, so we trap people failing to specify the build type with a fatal error. To fix, tell cmake what build type you want e.g.
I was just about to say "we tell you this in the build docs", but when I examine https://github.com/ned14/llfio/blob/develop/Build.md, I realise we left out the cmake build type. So you raise a completely valid documentation bug, for which I apologise, and I'm changing the name of this issue to match the documentation bug. Sorry about the bad documentation! |
ned14
changed the title
|
Thanks, it now builds! I also wanted to run the tests, and I got the error for "ctest -R llfio_sl". However, both "ctest -C Release -R llfio_sl" and "ctest -C Debug -R llfio_sl" worked, and resulted in failures of This seems to indicate to me that the last line of the build instructions should also be changed accordingly. Here also the output for the failed tests if that is in any way interesting: Output of "ctest -C Release -R llfio_sl --rerun-failed --output-on-failure" Output of "ctest -C Debug -R llfio_sl --rerun-failed --output-on-failure": |
|
The extended attributes test failing is surprising. What's your filing system and did you mount it with extended attributes disabled? The path view tests failing is also surprising given how recent your libstdc++ is, and the fact these tests pass on GCC on CI. BTW, path_view is correct above, your standard library is not. The current process memory usage test can be a touch spurious if anything else is happening on the machine in the background. It's fine on the CI runners. It's hard to make it not spurious if you want it to actually test anything real. Thanks for this feedback, it's always useful to discover other people's experiences. |
|
I just checked, and yes, user attributes are disabled, here the output of mount: (this is a default btrfs filesystem created from a vanilla installation of Fedora when erasing the whole disk) fstab is also default generated: |
ned14
added a commit
that referenced
this issue
Dec 16, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I tried to build LLFIO as a static library on Fedora 37 (current Clang and GCC) and got the following output on "cmake ..":
The content of CMakeOutput.log is:
Clang version info:
GCC version info:
My entered commands were (in the fish shell, if that's somehow important):
The text was updated successfully, but these errors were encountered: