FTBFS (src/ai/defaultai.cc:3898:1: error: control may reach end of non-void function)
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
widelands |
Fix Released
|
Medium
|
Unassigned |
Bug Description
Current trunk now fails to build with the following error message:
/«PKGBUILDDIR»
For details, see https:/
Presumably, this is because the assert() is trimmed out from release builds which means the else doesn't contain anything. Might be resolved/handled by throwing an exception saying something like "unhandled purpose", though I don't know the policy on exceptions in such cases.
Also, I've had general issues building widelands with clang 3.6 and later, but this has been going on for a while. I haven't had time to properly investigate, but once this is resolved I might take another look at it. Looked like it simply had problems including some of the SDL-libraries for some reason, has anyone else run into similar problems?
Related branches
- TiborB: Disapprove
- SirVer: Approve
- GunChleoc: Approve
-
Diff: 11 lines (+1/-0)1 file modifiedsrc/ai/defaultai.cc (+1/-0)
summary: |
- FTBFS with clang (src/ai/defaultai.cc:3898:1: error: control may reach - end of non-void function) + FTBFS (src/ai/defaultai.cc:3898:1: error: control may reach end of non- + void function) |
description: | updated |
tags: | removed: clang |
Changed in widelands: | |
status: | New → In Progress |
Changed in widelands: | |
status: | In Progress → Fix Committed |
Changed in widelands: | |
status: | Fix Committed → Fix Released |
Tibor has already crated a patch for the first issue.
I have noticed some build issues on our experimental Travis as well. Over there, they seem to be related to the ICU - some enum members I'm using aren't available, which points to the assumption that the ICU version over there needs to be newer. I guess we need to edit a requirement for the latest version in Cmake somewhere. The current stable ICU version listed on on their site is 56.1.