Highly portable C system library: threads and synchronization primitives, sockets (TCP, UDP, SCTP), IPv4 and IPv6, IPC, hash functions (MD5, SHA-1, SHA-2, SHA-3, GOST), binary trees (RB, AVL) and more. Native code performance.
License: MIT License
Some APIs could be detected at runtime instead of the configuration process. Use runtime API detection if possible.
For now only GCC 4.x is used. We should also build the library with more toolchains, at least GCC 5 & 6.
There is a PPA with toolchains for Ubuntu Precise: https://launchpad.net/~ubuntu-toolchain-r/+archive/ubuntu/test
Installing packages on Travis: https://docs.travis-ci.com/user/installing-dependencies/
As different compilers may require different flags as well as there are several atomic models in the wild it should be clarified how to properly configure it.
Some platforms (i.e. BeOS) don't have a last error mechanism for shared library loading. That's why sometimes we need to store error information directly within the library handler. To do that we need to change API for p_library_loader_get_last_error() to pass a library handler.
OpenVMS is a proprietary operating system currently developed by HP and VMS Software. It supports VAX, Alpha and IA64 architectures. I have made some preliminary research and work on possible OpenVMS port with the result described below.
VAX platform is not a player for several reasons:
Alpha and IA64 platforms are fully supported and new versions are released almost simultaneously.
The main challenge for OpenVMS port is a non-UNIX-like environment of the system. Though it has GNV (GNU-like environment with bash and friends), it doesn't simplify porting great. I have seen some threads with attempts to build CMake on OpenVMS, but none of them was completely successful. Anyway I do not like GNV approach, that why we need the following things to be done for the OpenVMS port:
bash-like command line interpreter on OpenVMS);What I have already done:
Almost none of that work was committed and still requires to be polished, but it is not a bad start. All tests were done on an Alpha emulator, but I can check it on real hardware later. IA64 platform still needs to be evaluated more closely.
Now all the private routines are collected in the single header file. It will be better to split it to several independent header files.
It would be nice to check and add Clang compiler support for Windows.
A new API should be able to obtain the library version in the numeric form as well as in the string one.
Add support for spinlocks based on atomic operations. It should provider a light-weight alternative for mutex-based locking.
7.1.3 Reserved identifiers
Each header declares or defines all identifiers listed in its associated subclause, and optionally declares or defines identifiers listed in its associated future library directions subclause and identifiers which are always reserved either for any use or for use as file scope identifiers.
— All identifiers that begin with an underscore and either an uppercase letter or another underscore are always reserved for any use.
— All identifiers that begin with an underscore are always reserved for use as identifiers with file scope in both the ordinary and tag name spaces.
It includes SHA-2/384 and SHA-2/512 algorithms.
Current thread implementations lack thread local storage (TLS) support.
Add API to receive self thread ID.
For now little endian mode is used by default and there is no way to select big endian mode. OpenVMS on Alpha and Itanium can be configured for big endian mode. Thus we should have an option for build_openvms.com script to select an endianness.
Every thread stores its data structure in a TLS cell. We should add API to retrieve that data structure for the current thread.
p_uthread_create_full() should accept the additional stack size parameter.
Recently PGI compiler became available for free (community edition).
Library and test objects should be located in separate directories to not to mix with each other while creating .OLB library.
For CPU architecture, OS and compiler detection macros Doxygen generates duplicates for the target host.
It is better to give a URL to license text than mentioning a postal address.
Add library initialization routine with memory table.
We can detect OS and compiler, but still no way to detect CPU architecture. Thus, we can't target software for a specific hardware architecture. A set of CPU detection macros is required.
Windows and POSIX implementation reside in a single source file. For a better flexibility they should be split. Also a new corresponding model should be available for platform configuration.
HPPA-RISC32 architecture doesn't support reference count in shl_unload(). Thus, it unloads all references (including our own), which leads to a test crash. To prevent this behaviour, we need to detect HPPA-RISC32 architecture and adopt the test properly.
This issue depends on #44.
There is LLVM 3.9.0 (x64) available on a worker machine. It can be combined with MSVC 2015.
Can be downloaded here: https://sourceforge.net/projects/openwatcom/files/open-watcom-1.9/
On AppVeyor GCC 7.x builds from MinGW-w64 are already available.
On Travic-CI GCC 7.x builds should be available at least for macOS images using Brew.
We need to extent threading API with ability to get to know how many CPU core are available for optimal performance.
Implement RW-lock thread synchronization primitive.
All tests are done using almost normal situations with invalid input parameters. We can enhance testing with memory allocation failures using PMemVTable with fake memory management calls.
On HP-UX prior to 11.xx dlopen() and its family functions may not be implemented. Instead there is a family of HP-UX specific calls shl_*.
HP-UX manual says that developer should use dlopen() family, so we need to test its presence and choose an appropriate implementation model.
Basic support should be provided:
Current imlementation is working only on a limited number of POSIX-compliant systems. New implementation should dynamically calculate priority based on system capabilities.
When errno is set to EVMSERR the value of the vaxc$errno variable should be used for the native error code.
A new version of QNX 7 SDP is available. It has 30-days evaluation period, thus can be used for testing builds also.
The implementation should provide 224, 256, 384 and 512 bit variants.
Solaris has its own code coverage tool named tcov.
About using tcov: Analyzing Program Performance with Sun WorkShop.
About -xprofile option for Oracle/Sun compiler: Oracle docs.
I have also found some useful information about Oracle/Sun compiler versions: old versions and newer versions. We need it because there two ways to enable code coverage, and according to some threads a new way was supported at least since SunPro C 4.x version.
During library configuration we should select one of the code coverage ways.
clock_nanosleep() allows to chose CLOCK_MONOTONIC clock source, which can help to optimize power saving. It also easier to avoid time drifting problem (repeated interruption by a signal) when using TIMER_ABSTIME flag.
Some documentation need to be updated, some to be rewritten.
Simply looping ans checking for the EINTR return code may lead to weird things in multithread environment. You can double-close a socket descriptor: the first time your own, and the next time allocated for another thread (with the same descriptor value). The problem is specifically unpleasant on some systems which may return with EINTR without actually closing a socket descriptor.
There is no general way to handle the close() call on all operating systems, so we need to do some checks.
Some related links:
http://alobbs.com/post/54503240599/close-and-eintr
https://bugs.chromium.org/p/chromium/issues/detail?id=269623
http://austingroupbugs.net/view.php?id=529
http://www.daemonology.net/blog/2011-12-17-POSIX-close-is-broken.html
It includes SHA-2/224 and SHA-2/256 algorithms.
Currently GCC 6.2.x is used, but 6.3.x builds are already available.
These macros can improve performance for the branching code.
OS/2 is a proprietary operating system which supports x86 hardware architecture. Despite being quite old it is still in use and there are some (compatible) reincarnations (eComStation, Blue Lion) with ongoing support and new features. So, overall, the system is alive and it would be nice to support it.
The amount of work depends on the way we choose:
First way is definitely not a deal. Second one lets work to be done quickly using emulated POSIX environment, but some of the required parts are still not available (pthreads, mmap). The last one should be the most tricky and requires more efforts, but is gives full control over the situation and shouldn't impact on overall performance.
As we should avoid additional external dependencies like kLIBC, it will be better to start with a native way and go with IBM's Developer Toolkit 4.5.x (provides headers and import libraries).
Anyway, we need development environment with bash, make and CMake, which are freely available. There are several compilers for OS/2 (GCC, Watcom, VisualAge), but we should start with GCC as one of the most popular.
To port the library we need to implement the following models:
_beginthread() and friends from C runtime library) and synchronisation primitives.mmap() with something like DosAllocMem().DosAllocSharedMem() along with a process-wide mutex.DosTmrQueryTime () or DosGetDateTime() more closely).So, on the first impression, it is possible to make a native port to OS/2, maybe not feature-complete, but with most of them.
Some operating system are similar to UNIX-like, but actually they don't belong to UNIX/BSD families. For such system P_OS_UNIX macro should not be defined.
It will be better not only to print the module name in which an error occurred, but the routine name as well.
There should be an option like PLIBSYS_TESTS which could disable attempts to find Boost and build tests.
Tru64 platform page is missing.
Currently scope ID and flow info are not explicitly supported by API (partially some OS can handle given addresses properly). We need to use getaddrinfo() when converting string address to native structure. Also it would be nice to have an ability to read those attributes from the socket address object.
_beginthreadex() should be used to prevent from memory leaking while using CRT libraries.
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.