Empty docs in Xcode 11 (despite latest sourcekitten) #643
Comments
|
It doesn't look like you're doing anything particularly unusual. It is odd that every time, sourcekitten goes down the "Checking xcodebuild -showBuildSettings" path: this means that it thinks the regular 'xcodebuild ' did not build your module and so it goes digging around for a build record. I don't understand this because you're explicitly cleaning the build directory, so there's no reason for the regular command to not build the module. I guess you cannot share the project; if you can spare time to debug then these are the commands sourcekitten looks to be running: This is for use later but does seem to create the temp directory. ...this is the one in whose output it can't find a Now grab Now look at We'll use these Your data suggest that, in the failure case, sourcekitten is successfully finding this This kind of problem has been caused in the past by bad handling of non-basic characters in pathnames/module names/etc, so if there's anything like that in your system it would be worth noting. I am moderately confident that we've sorted out spaces/regular shell-escapes though. |
|
@johnfairh Thanks for the quick reply. Is there an easy way to run I just realised that while using Cmd-B in Xcode (on the same project and same scheme as provided to (invocation with |
I was thinking this would be a good thing as i wrote the reply! The rest of your comment explains my first confusion. Perhaps the build is failing in a way to correctly generate the build record but not the contents of the response file - explaining why Xcode 10 and |
|
Ahah! I think I found the culprit, thanks to your steps:
But that So it was trying to build for Catalyst instead of iOS after all 🤦♂ A verbose mode, or at least even printing the |
|
Not sure what's the relationship between "using How come adding I mean, the issue is solved for our case and project, adding |
|
That's great. Remember that sourcekitten's So, it looks like the What was the error from |
Here's the output of It starts normally[snip] And a lot more SymLink and MkDir, omitted for brevity [/snip] [snip] And a lot more WriteAuxiliaryFile on yaml/hmap/modulemap files, omitted for brevity [/snip] [...] [snip] And some more ProcessInfoPlistFile, CpHeader and Ditto, omitted for brevity [/snip] Then we get to the CompileC invocations[snip] Some other CompileC invocations that seem to succeed, omitted for brevity [/snip] Then at some point we encounter some MacCatalyst errors on CompileSwiftSources[snip] Then some more CompileC invocations without error [/snip] And finally some nonzero exit code on CompileSwift invocations[snip] Some other CompileSwift invocations without nonzero exit codes [/snip] And finally the "BUILD FAILED" summaryPS: Obviously, otoh running the same command but with |
|
Ok, thanks for all that. Probably significant that it fails during dependencies, not got round to building your framework. I guess best to keep this issue open to better report xcodebuild failures. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hello,
We are using
sourcekitten 0.29.0, but we encounter the same issue that the one described in realm/jazzy#1108 … despite that issue supposedly being fixed by #618 and the fix being shipped in0.26.0Not sure if we're holding it wrong, or if the issue was just not entirely fixed by #618 in the first place at we hoped…
Any idea @jpsim @johnfairh ?
Empty docs with Xcode 11, despite using sourcekitten 0.29 which is supposed to have that fix:
Fix to pass
USE_SWIFT_RESPONSE_FILE=NOsuggested in realm/jazzy#1108 works (but feels like a workaround, especially now that this is supposed to be fixed since 0.26):Using Xcode 10.3 (via
DEVELOPER_DIR=…) works like it used to, even without theUSE_SWIFT_RESPONSE_FILE=NOworkaround. So issue is only present since we migrated to Xcode 11:The text was updated successfully, but these errors were encountered: