Bug 1306226

Created attachment 1122748 [details]
build.log

Created attachment 1122749 [details]
root.log

Created attachment 1122750 [details]
state.log

OBULLET.cpp: At global scope:
OBULLET.cpp:41:78: error: narrowing conversion of '-1' from 'int' to 'char' inside { } [-Wnarrowing]
  { 0, 0,-1, 0, 1,-1,-1, 1, 1, 0,-2, 0, 2, -1,-2,-2,-1, 1, 2, 2, 1,-2,-2, 2, 2};

That's cause gcc6 enforces you to use 'signed int' instead of just 'int' as the variable type. We've seen that already in dreamchess-tools, psi and iris packages. A patch should be easy to do and upstreamable.

Odd. I can not reproduce locally for f24 with fedpkg mockbuild, but fedpkg scratch-build fails on koji.

> That's cause gcc6 enforces you to use 'signed int' instead of just 'int' as the
> variable type.

signed char actually. int is always signed. char can be signed or unsigned depending on the platform. On ARM, it defaults to unsigned. (You can use the -fsigned-char switch to force it to be signed, but the code should use "signed char" explicitly for maximum portability.) This previously compiled without error, but did not necessarily work right, because -1 was silently converted to 255 on ARM.

We need a proper upstream(able) patch. Unfortunately, I do not see reasonable spare time to do that by myself in the near future.

This bug appears to have been reported against 'rawhide' during the Fedora 24 development cycle.
Changing version to '24'.

More information and reason for this action is here:
https://fedoraproject.org/wiki/Fedora_Program_Management/HouseKeeping/Fedora24#Rawhide_Rebase

Still b0rken. Maintainer, please urgently give any feedback or orphan this package.

It would be really sad to see this great game go away again in Fedora, I'm wondering why I did the package review recently (for you) and you can not actively care any more about package consistency as it's required by package maintainer responsibilities in our guidelines.

I intended to fix it around later this week or early next week.

Friendly reminder (again). Please notice the awareness of FeSCo and take care to not get the package orphaned.

yselkowitz's 7kaa-2.14.5-12.fc25 completed http://koji.fedoraproject.org/koji/buildinfo?buildID=742816

The assumption that char is signed exists throughout the code.  A patch fixing all such occurrences would be quite extensive.  As a leaf package, building with -fsigned-char is the simplest solution.

Wouldn't it be better to request from upstream the general change to unsigned char?

(In reply to Raphael Groner from comment #14)
> Wouldn't it be better to request from upstream the general change to
> unsigned char?

signed char is fine, it just needs to be explicit.  Feel free to work with upstream for a fix in the code itself, but I looked into this and the changes would be extensive.

(In reply to Yaakov Selkowitz from comment #15)
> (In reply to Raphael Groner from comment #14)
> > Wouldn't it be better to request from upstream the general change to
> > unsigned char?
> 
> signed char is fine, it just needs to be explicit.  Feel free to work with
> upstream for a fix in the code itself, but I looked into this and the
> changes would be extensive.

Ack. Besides the fact many upstreams do not care that much about ARM as an important platform to develop on, especially in gaming sector (my experience so far).

Overseen you built for f24 as well, sorry.