octave-bug-tracker
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Octave-bug-tracker] [bug #60567] oct-file has an undefined symbol


From: Markus Mützel
Subject: [Octave-bug-tracker] [bug #60567] oct-file has an undefined symbol
Date: Tue, 11 May 2021 09:58:22 -0400 (EDT)
User-agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/90.0.4430.93 Safari/537.36 Edg/90.0.818.56

Update of bug #60567 (project octave):

                  Status:                   Fixed => In Progress            
             Open/Closed:                  Closed => Open                   

    _______________________________________________________

Follow-up Comment #8:

Tbh, I don't understand completely how visibility is implemented by different
compilers and vendors on different platforms. It's mostly trial and error for
me.
I saw that adding those flags didn't break anything with gcc and clang on
Ubuntu, and cross-compiling with gcc for Windows still worked. They also seem
to have fixed the motivating undefined symbol issue. So, I went ahead and made
those changes. (It might have been better to post the patch here and wait for
feedback.)
The changes also seem to have allowed proceeding a little bit further in the
build process with Apple clang on macOS (see bug #59820).

But I agree that it might be preferable to try to apply those flags more
minimalisticly.

This change in CRowVector.h might have sufficed to resolve the motivating
error:

-  friend ComplexRowVector operator * (const ComplexRowVector& a,
-                                      const ComplexMatrix& b);
+  friend OCTAVE_API ComplexRowVector
+  operator * (const ComplexRowVector& a, const ComplexMatrix& b);
 
-  friend ComplexRowVector operator * (const RowVector& a,
-                                      const ComplexMatrix& b);
+  friend OCTAVE_API ComplexRowVector
+  operator * (const RowVector& a, const ComplexMatrix& b);


It involves friend operator functions like you wrote. Should we add the flags
only to friend functions?

I don't mind at all reverting 3e8bc8bee8e9.

Opening a discourse thread might be a good idea. It sounds like you already
know what to do. But we can still ask for further input if you prefer.


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?60567>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/




reply via email to

[Prev in Thread] Current Thread [Next in Thread]