eix 0.33.2 uses std.vector<bool>.emplace_back which is a c++14 feature. (http://en.cppreference.com/w/cpp/container/vector_bool) But it sets CXXFLAGS to -std=c++11 Reproducible: Always Steps to Reproduce: 1.set used compiler to clang with 'default-libcxx' Actual Results: fails compile with: cache/parse/parse.cc:382:17: error: no member named 'emplace_back' in 'std::__1::vector<bool, std::__1::allocator<bool> >' further_works.PUSH_BACK((*it)->readCategoryPrepare(cat_name)); ~~~~~~~~~~~~~ ^ ./eixTk/dialect.h:49:19: note: expanded from macro 'PUSH_BACK' #define PUSH_BACK emplace_back Expected Results: it compiles
emplace_back is a c++11 feature. However, eix has a ./configure test which checks whether the feature can be used. I am wondering why this test succeeds while the actual compilation fails. Does it work if you replace in line 303 and 304 MV_APPEND by MV_PREPEND (and then run ./autogen.sh or eautoreconf to regenerate ./configure)?
Whoops, wrong patch suggestion... The patch I meant was to replace MV_APPEND by MV_PREPEND in line 942 (and only there).
i think the error is somewhere else. if i execute ./configure manually it gets c++17. ebuild eix-0.33.2.ebuild configure --debug reports at the end: CXX: clang++ CXXFLAGS: -march=native -O2 -pipe -std=c++11 PREPEND_CXXFLAGS: -fdata-sections -ffunction-sections LDFLAGS: -Wl,-O1 -Wl,--as-needed PREPEND_LDFLAGS: -Wl,--gc-sections i think these lines: + local myconf + append-cxxflags -std=c++11 + [[ 1 -eq 0 ]] + export 'CXXFLAGS+= -std=c++11' + CXXFLAGS+=' -std=c++11' + return 0 + econf --disable-paranoic-asserts --enable-nls --without-extra-doc --with-sqlite --without-always-accept-keywords --with-dep-default --with-required-use-default --with-portage-rootpath=/usr/x86_64-pc-linux-gnu/gcc-bin/7.3.0:/usr/lib/llvm/5/bin:/usr/local/sbin:/usr/local/bin:/usr/sb in:/usr/bin:/sbin:/bin:/opt/bin --with-eprefix-default= --disable-separate-binaries --disable-separate-tools --disable-debugging --disable-new_dialect --disable-optimization --disable-strong-optimization --disable-security --disable-nopie-security --disable-strong-security point to the cause. the ebuild states in line 76-77: # https://github.com/vaeth/eix/issues/35 append-cxxflags -std=c++11
(In reply to therjak from comment #3) > if i execute ./configure manually it gets c++17. Unless you specify --disable-new_dialect (which is meanwhile not the default anymore) ./configure chooses the most current available dialect. In any case, there is a ./configure check whether emplace_back is working with your CXXFLAGS (and possibly appended dialect flags from ./configure which finally will be prepended to your CXXFLAGS): If ./configure finds that it is not working with your compiler (for whatever reason) with these flags, emplace_back is not used. So the only possible difference between the ./configure test and the actual compilation I can imagnine is that in the former case the flags are appended while in the latter the flags are prepended. If this is not the reason, it might help to see the configure output (to see whether the test indeed succeeds) and the config.log output (if the test succeeded which flags were used in contrast to the final compilation.)
configure checks if emplace_back works on vector<pair<int,int>> and set<pair<int,int>> which are c++11 feature. emplace_back on vector<bool> is a c++14 feature and is not checked.
Thank you for the clarification. I was not aware of an exception in the standard for vector<bool>. (Also, gcc-7 does not seem to make such a difference.) I have fixed the configure test in eix correspondingly to include vector<bool>; I will not write special code only to optimize for c++11 but not c++14. So passing -std=c++11 (which I did not recommend anyway, see the cited link) should "just" force non-optimizing emplace_back with >=eix-0.33.3
*** Bug 654360 has been marked as a duplicate of this bug. ***
@Martin, I still have the same issue @therjak pointed out std::vector<bool>::emplace and std::vector<bool>::emplace_back are c++14 features and I have the build failed Is something being planned to take the issue down? Why wouldn't you like to set c++14 rather than c++11 explicitly? There are almost no essential differences in terms of working ability of code, just smallish improvements relatively to c++11 Thank you
I am not a proxied maintainer. The eix build system does set correct -std according to the compiler's ability. The eix ebuild in the gentoo repository intentionally overrides that build system against my explicit recommendation. The background is political or perhaps personal (see the discussion preceding commit 39f0ae3fe1259d17abc10cc0b21e2c9820089292 if it should still be on github); changes related to that commit are politically dangerous for gentoo developers. I recommend to use the ebuilds from the mv overlay for my packages: These do not suffer from politics, and bugs are fixed immediately.
So does that make this a WONTFIX?
> So does that make this a WONTFIX? A few minutes ago, I released eix-0.33.5, so from my side the bug was already fixed 3 releases ago: The eix bug (that compilation with pure c+11 dialect does not work) is fixed since eix-0.33.3. The fix is not optimal (for simplicity, all emplace_back-optimizations are avoided if vector<bool>::emplace_back does not work), but this is only a minor optimization issue; the non-usage of other optimal CFLAGS for eix in the gentoo ebuild has other much worse speed and memory penalties anyway.
(In reply to Sora Lee from comment #10) > So does that make this a WONTFIX? The maintainer could just swap c++11 for c++14. However, since he didn't reply to this bug for 6 months, I'm going to claim maintainer timeout and fix it.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=f8039025292b2bbac64eedc22c6321ac442559fb commit f8039025292b2bbac64eedc22c6321ac442559fb Author: Michał Górny <mgorny@gentoo.org> AuthorDate: 2018-09-29 08:06:58 +0000 Commit: Michał Górny <mgorny@gentoo.org> CommitDate: 2018-09-29 08:16:48 +0000 app-portage/eix: Fix C++ standard to C++14 to fix code bugs Closes: https://bugs.gentoo.org/649558 Signed-off-by: Michał Górny <mgorny@gentoo.org> app-portage/eix/eix-0.33.2-r1.ebuild | 109 +++++++++++++++++++++++++++++++++++ 1 file changed, 109 insertions(+)
