kdevelop build error: undefined reference to `KDevelop::EditorContext::currentWord() const'

[Aetf]

On Sat, Jun 25, 2016 at 8:56 AM René J.V. <rjvbertin at gmail.com> wrote:

> Do you have the possibility to test with an earlier version, gcc 5.x or
> even 5.4?
>

That's the only version of gcc I have right now. Arch Linux made the
movement from 5.x series to 6.x recently and it's not easy to use a earlier
version after that. But I have clang 3.8, will try that and see if there's
any difference.


> >Oops, actually `#include <signal.h>` also works on OS X. It must be me
>
> The manpages for OS X are available online just like they are for Linux ;)
>

Sorry, I meant `#include <signal.h>` also works on *Linux* :P.


> I'm seeing -fvisibility=hidden -fvisibility-inlines-hidden but they're set
> automatically and are used almost everywhere. Apparently not in
> debuggers/gdb, though.
>

That may in fact explain the issue, maybe debuggers/gdb contains something
> that causes it to be built with default (public?) visibility and you then
> moved things to debuggers/common without ensuring that part is built with
> the same visibility rules.
>

The code I moved is built with the same visibility as kdevgdb target. The
thing is, files in debuggers/gdb are actually compiled twice, once for
target kdevgdb, once for target test_gdb. And -fvisibility argument was not
used when compiling for test_gdb and causes it to be built with default
visibility, while all other parts are compiled with hidden visibility.

But I can't find any rules in CMakeLists.txt that are changing visibility
settings for test_gdb target, maybe it's somewhere in ecm_add_test?


> But that wouldn't explain why EditorContext::currentWord() and the
> EditorContext typeinfo are the only symbols concerned...

R.
>

No idea why is that either...

Aetf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kdevelop-devel/attachments/20160625/d61457c4/attachment.html>