sci-physics/reduze: Fix building with GCC-6 #4796
Conversation
Bug: https://bugs.gentoo.org/show_bug.cgi?id=594900 Package-Manager: Portage-2.3.6, Repoman-2.3.2
| ex resultLeadingNc = result; | ||
| for (exset::const_iterator it = found.begin(); it != found.end(); ++it) { | ||
| - VERIFY(is_a<function>(*it) && ex_to<function>(*it).get_name() == "Color"); | ||
| + VERIFY(is_a<GiNaC::function>(*it) && ex_to<GiNaC::function>(*it).get_name() == "Color"); |
There was a problem hiding this comment.
This change is already in reduze-2.1.
gentoo-repo-qa-bot
added
the
assigned
|
Pull Request assignment Areas affected: ebuilds sci-physics/reduze: |
|
😞 The QA check for this pull request has found the following issues: Issues inherited from Gentoo (may be modified by PR): |
| get_filename_sectormappings(m.integralfamily()->name()); | ||
| ofstream file(file_name.c_str()); | ||
| if (!file) | ||
| - ABORT("Can't write to " << file); |
There was a problem hiding this comment.
In C++98, streaming a stream object would have the effect of casting it to a void * and streaming its address (or NULL address if the failbit was set). This was omitted in C++11. IMHO, he best practice for debugging purposes is to stream the name of the file instead ("Can't write to test.txt" instead of "Can't write to 0x000000000000").
There was a problem hiding this comment.
or just addressof it?
ABORT("Can't write to " << &file_name);
| bool is_readable_file(const string& filename) { | ||
| ifstream is(filename.c_str()); | ||
| - bool ok = is; | ||
| + bool ok = !is.fail(); |
There was a problem hiding this comment.
Again, is would be casted to void * and then to bool in C++98. !is.fail() has the same effect and is dialect neutral.
| for (Iterator n = z_node.begin(); n != z_node.end(); ++n) | ||
| for (Iterator s = n.second().begin(); s != n.second().end(); ++s) | ||
| - m.zero_sectors_.insert(*s); | ||
| + m.zero_sectors_.insert((const int) *s); |
There was a problem hiding this comment.
This calls std::set::insert() on YAML::Node::operator int() const but in C++11/14 it seems to cause an ambiguity between insert( const value_type& value ), insert( value_type&& value ), and insert( std::initializer_list<value_type> ilist ). Explicitly casting it to const int works to overcome this and is dialect neutral.
Bug: https://bugs.gentoo.org/show_bug.cgi?id=594900
Package-Manager: Portage-2.3.6, Repoman-2.3.2
No place for bug submissions or PRs upstream.
Note: Both ebuilds fail when built against 'sci-mathematics/ginac-1.7.2'. This is an unrelated bug.