- 02 Aug, 2018 3 commits
-
-
Artem Shinkarov authored
Some of the tests do not pass, as there are bugs in the implementation of string functions.
-
Artem Shinkarov authored
-
Artem Shinkarov authored
With the use of gtest, we will write tests in C++, so we could have specified C, CXX as languages, but I don't see any reason to do this.
-
- 30 Jul, 2018 1 commit
-
-
Hans-Nikolai Viessmann authored
update HWLOC dependence for DEB package See merge request !54
-
- 28 Jul, 2018 1 commit
-
-
Hans-Nikolai Viessmann authored
Use stdbool.h for the bool type definition. See merge request !56
-
- 27 Jul, 2018 2 commits
-
-
Artem Shinkarov authored
-
Artem Shinkarov authored
-
- 09 Jul, 2018 1 commit
-
-
Hans-Nikolai Viessmann authored
The previously given dependency for the DEB packages was incorrect, as it did not install the headers - only the runtime library of hwloc.
-
- 27 Jun, 2018 1 commit
-
-
Hans-Nikolai Viessmann authored
CPACK fix (HWLOC dependency missing) See merge request !53
-
- 26 Jun, 2018 2 commits
-
-
Hans-Nikolai Viessmann authored
also made minor fix cpack config
-
Hans-Nikolai Viessmann authored
-
- 14 Jun, 2018 4 commits
-
-
Hans-Nikolai Viessmann authored
Artem fuzzer fixes See merge request !51
-
Artem Shinkarov authored
-
Artem Shinkarov authored
-
Artem Shinkarov authored
Otherwise, grep produces som nonsense.
-
- 12 Jun, 2018 7 commits
-
-
Artem Shinkarov authored
-
Artem Shinkarov authored
-
Artem Shinkarov authored
It is surprising, but we can define a ++/-- function with no parameters and call it is a normal function. The code that was handling the ++/-- (postfix operator resolution) was making an assumption that there is one argument.
-
Artem Shinkarov authored
-
Artem Shinkarov authored
-
Artem Shinkarov authored
A program to cause a failure: int [ain,() { return 0; } The rule for a type is "<base-type> ('[' <expr-list> ']')?". When expr-list was parsed, an erroneous component was turned into NULL, and then a function was trying to set location to the NULL node. We check now that the component is not null, as "<base-type> '[' ']'" is not supported by the sac syntax.
-
Artem Shinkarov authored
An example to reproduce the failure is this: int [ain () { return 0; } The reason for the failure was that error_type_node was propagating into a TBmakeRet type constructor and caused segfault when trying to free the node. The patch fixes the problem and simplifies handling of return types.
-
- 11 Jun, 2018 2 commits
-
-
Artem Shinkarov authored
[vers manager] minor fix to support python2.6 See merge request !50
-
Hans-Nikolai Viessmann authored
On CentOS 6.*, Python2.6 is the primary variant of python available through the package manager. Unfortunately, it is very cumbersome to install a newer version of Python, as several systems tools (such as yum) have a hard dependency on the older version. Because of this we run into compatibility problems: Python2.6 does not support dict comprehensions (these were introduced in python2.7 - see [PEP274][1]). We use this in two places within the version script. This patch changes the comprehensions into normal loop operations. This works fine with later versions of Python2.* branch. It's unfortunate that we need to worry about this given that python2.6 is no longer support by the Python devs... perhaps we should rethink out position on supporting CentOS 6. [1]: https://www.python.org/dev/peps/pep-0274/
-
- 06 Jun, 2018 1 commit
-
-
Hans-Nikolai Viessmann authored
fixed OSX problem of parameterised CC flag See merge request !48
-
- 30 May, 2018 1 commit
-
-
Sven-Bodo Scholz authored
OSX wants to add something like -isysroot <BLA> which needs to be handed through. However, -Xc and -Xl do not support flags with parameters. The solution here is to transform this into -Xc -isysroot -Xc <BLA>. This is not ideal but all other attempts to pass an argument failed :-(
-
- 25 May, 2018 7 commits
-
-
Hans-Nikolai Viessmann authored
Artem remove old buildsystem Closes #1212 See merge request !45
-
Artem Shinkarov authored
This helps to avoid rebuilding the entire project when entering subdirectories, and it makes -j flag to work correctly on subdirs.
-
Artem Shinkarov authored
-
Artem Shinkarov authored
-
Artem Shinkarov authored
-
Artem Shinkarov authored
-
Artem Shinkarov authored
-
- 17 May, 2018 2 commits
-
-
Hans-Nikolai Viessmann authored
GCC 8 warning fixes [unsure] See merge request !46
-
Hans-Nikolai Viessmann authored
Thie commit resolves some of the new warnings that GCC 8 issues. Specifically warnings `-Wcast-function-type` and `-Wrestrict`. The first clearly deals with casting from one function type to another and the latter deals with breaks in `restrict`, that is we have _unsafe_ operations, e.g. reading and writing using the same pointer.
-
- 11 May, 2018 1 commit
-
-
Artem Shinkarov authored
Extend sac2c-version-manager to support installing symlinks in a more generic way See merge request !44
-
- 04 May, 2018 2 commits
-
-
Hans-Nikolai Viessmann authored
-
Hans-Nikolai Viessmann authored
In particular the install command documentation has been changes to be more clear. Other commands have also been changed slightly. Additionally we've cleaned up the code a bit and have included some more CMake variables from the configuration phase in the script - we used to infer these but it is likely better to rely on the build system for this information.
-
- 03 May, 2018 2 commits
-
-
Hans-Nikolai Viessmann authored
This is needed for the sac-version-manager script. We could check for version 2.7 (where argparse is include per default), but CentOS has a hard dependency on Python 2.6.6 (which doesn't have argparse). There argparse needs to be setup seperately.
-
Hans-Nikolai Viessmann authored
Previously, we used `cmake -E` internal symlink mechanism to create links. We now use the sac2c-version-manager script instead, as this is slightly more reliable when it comes to 'installing' different links. This is also multi-platform compatible, though for Windows and forwards slashes this is untested - Python os.* supports Windows so it should work, or otherwise gracefully fail.
-