-
-
Notifications
You must be signed in to change notification settings - Fork 55
GCC 7.0.1, OpenCoarrays 1.8.10 build failure on FreeBSD #388
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
The log doesn't show:
Would it be possible to include this information? It's strange, because |
working on your questions. |
works to build 1.8.10: http://eis.bris.ac.uk/~mexas/opencoarrays-1.8.10-CC=gcc7-make.log |
Regarding the invocation of cmake, isn't this what you are looking for
|
By the way, is opencoarrays build thread-safe?
|
Yes, unless you do something funky when patching it (debian's patch builds static+dynamic libs, and didn't engineer the patch to be thread safe... we're incorporating a thread safe version of their patch soon...)
No, what I am looking for looks like
It also apparently includes the option
Great! As I mentioned before, AppleClang 7.0.2.7000181 doesn't seem to have any issues building OpenCoarrays. My suspicion is that your Clang is doing something wrong or funky. The bottom line is that something went wrong with the includes, since compiling with Clang, the error is:
BUT, It's possible that the this might be able to be fixed with a change to the build system (i.e. if the order of the include flags needs to be changed or by passing a full path to libcaf.h etc.) but I'm quite tempted to close this as "won't fix" since it works with GCC 7 and Clang on Apple. Regarding your patches:
|
Yes, I can build opencoarrays 1.8.10 with clang 5: I guess I need to send this to the clang team.
yes, ok to close this PR. Thanks |
Hi Anton, One final question: I notice that mpi_caf.c is also patched to remove If so I can add some introspection to the build system to see if it is present and if it is missing, remove it from mpi_caf.c. What do you think? |
On FreeBSD alloca is in libc: |
BTW, don't know why we removed hello world, |
Great, I hope to have a release out very soon with the alloca introspection
and experimental support for failed images (if you ask for it and compile
with MPI that has the proposed/experimental features required for failed
images support)
…On Tue, May 23, 2017 at 10:21 AM mexas ***@***.***> wrote:
BTW, don't know why we removed hello world,
might have been in the early stages of porting,
to narrow down the sources of build errors.
I removed that patch and all is well.
So we now have only the alloca patch on FreeBSD.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#388 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAREPOUO2i0KbE1y1u3lTdzvnqGR4Q03ks5r8utugaJpZM4NgwI7>
.
|
Once 151eb80 hits master/the next release, that last patch should no longer be required... |
- Test for existence, assume symbols/functionality provided elsewhere if it is missing. Should induce compile time error if this is not true.
Hi Anton, Can you please try the attached patch for the older clang and let me know how it goes? I added a little bit of debugging, so if it doesn't work, please upload the log again, and I may be able to get some more insight into why it's failing. From 4decd152dc9a492feca2d694785da4ea4cf50ecc Mon Sep 17 00:00:00 2001
From: Izaak Beekman <[email protected]>
Date: Tue, 23 May 2017 18:54:30 -0400
Subject: [PATCH] Patch *may* fix old clang issue, add new debuging
---
CMakeLists.txt | 2 +-
src/mpi/CMakeLists.txt | 4 ++++
2 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/CMakeLists.txt b/CMakeLists.txt
index f7f7b61..7ebc23b 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -383,7 +383,7 @@ include(GNUInstallDirs)
#-------------------------------
# Recurse into the src directory
#-------------------------------
-include_directories(${CMAKE_CURRENT_SOURCE_DIR}/src)
+include_directories(BEFORE ${CMAKE_CURRENT_SOURCE_DIR}/src)
add_subdirectory(src)
diff --git a/src/mpi/CMakeLists.txt b/src/mpi/CMakeLists.txt
index 6f4490b..a4feb5a 100644
--- a/src/mpi/CMakeLists.txt
+++ b/src/mpi/CMakeLists.txt
@@ -57,6 +57,10 @@ install(TARGETS caf_mpi EXPORT OpenCoarraysTargets
LIBRARY DESTINATION "${CMAKE_INSTALL_LIBDIR}"
)
+get_target_property(libcaf_mpi_includes caf_mpi INCLUDE_DIRECTORIES)
+message( STATUS "Include directories for libmpi_caf:\n\
+ ${libcaf_mpi_includes}")
+
# Install modules to standard include dir, but namespace them with compiler/version
set (mod_install "OpenCoarrays/${CMAKE_Fortran_COMPILER_ID}/${CMAKE_Fortran_COMPILER_VERSION}")
install(DIRECTORY "${CMAKE_BINARY_DIR}/mod/"
--
2.13.0
|
May fix #388 (issue with Clang 4 on FreeBSD)
just to confirm - 1.8.12 builds fine with no patches with clang 3.8.0. |
Ok, great!!! Just FYI I pushed the tag but have not created a release with
release assets. We're going to be releasing again with a minor version bump
and including better support for allocatable components very soon,
including release notes and release assets.
Does this mean we can close all of your outstanding FreeBSD issues? Is
there anything that remains unresolved?
…On Fri, May 26, 2017 at 6:20 AM mexas ***@***.***> wrote:
just to confirm - 1.8.12 builds fine with no patches with clang 3.8.0.
—
You are receiving this because you modified the open/close state.
Reply to this email directly, view it on GitHub
<#388 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAREPHF1CMhkz7zklNA4UAEvu_0W3botks5r9rWUgaJpZM4NgwI7>
.
|
Many tests still fail, I updated #348. |
full build log: http://eis.bris.ac.uk/~mexas/build.log
The error:
The text was updated successfully, but these errors were encountered: