docs/SystemLibrary.html

   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
   2                       "http://www.w3.org/TR/html4/strict.dtd">
   3 <html>
   4 <head>
   5   <title>System Library</title>
   6   <link rel="stylesheet" href="llvm.css" type="text/css">
   7 </head>
   8 <body>
   9
  10 <div class="doc_title">System Library</div>
  11 <ul>
  12   <li><a href="#abstract">Abstract</a></li>
  13   <li><a href="#requirements">Keeping LLVM Portable</a>
  14   <ol>
  15     <li><a href="#headers">Don't Include System Headers</a></li>
  16     <li><a href="#expose">Don't Expose System Headers</a></li>
  17     <li><a href="#c_headers">Allow Standard C Header Files</a></li>
  18     <li><a href="#cpp_headers">Allow Standard C++ Header Files</a></li>
  19     <li><a href="#highlev">High-Level Interface</a></li>
  20     <li><a href="#nofunc">No Exposed Functions</a></li>
  21     <li><a href="#nodata">No Exposed Data</a></li>
  22     <li><a href="#nodupl">No Duplicate Implementations</a></li>
  23     <li><a href="#nounused">No Unused Functionality</a></li>
  24     <li><a href="#virtuals">No Virtual Methods</a></li>
  25     <li><a href="#softerrors">Minimize Soft Errors</a></li>
  26     <li><a href="#throw">Throw Only std::string</a></li>
  27     <li><a href="#throw_spec">No throw() Specifications</a></li>
  28     <li><a href="#organization">Code Organization</a></li>
  29     <li><a href="#semantics">Consistent Semantics</a></li>
  30     <li><a href="#bug">Tracking Bugzilla Bug: 351</a></li>
  31   </ol></li>
  32 </ul>
  33
  34 <div class="doc_author">
  35   <p>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a></p>
  36 </div>
  37
  38
  39 <!-- *********************************************************************** -->
  40 <div class="doc_section"><a name="abstract">Abstract</a></div>
  41 <div class="doc_text">
  42   <p>This document provides some details on LLVM's System Library, located in
  43   the source at <tt>lib/System</tt> and <tt>include/llvm/System</tt>. The
  44   library's purpose is to shield LLVM from the differences between operating
  45   systems for the few services LLVM needs from the operating system. Much of
  46   LLVM is written using portability features of standard C++. However, in a few
  47   areas, system dependent facilities are needed and the System Library is the
  48   wrapper around those system calls.</p>
  49   <p>By centralizing LLVM's use of operating system interfaces, we make it
  50   possible for the LLVM tool chain and runtime libraries to be more easily
  51   ported to new platforms since (theoretically) only <tt>lib/System</tt> needs
  52   to be ported.  This library also unclutters the rest of LLVM from #ifdef use
  53   and special cases for specific operating systems. Such uses are replaced
  54   with simple calls to the interfaces provided in <tt>include/llvm/System</tt>.
  55   </p>
  56   <p>Note that the System Library is not intended to be a complete operating
  57   system wrapper (such as the Adaptive Communications Environment (ACE) or
  58   Apache Portable Runtime (APR)), but only provides the functionality necessary
  59   to support LLVM.
  60   <p>The System Library was written by Reid Spencer who formulated the
  61   design based on similar work originating from the eXtensible Programming
  62   System (XPS). Several people helped with the effort; especially,
  63   Jeff Cohen and Henrik Bach on the Win32 port.</p>
  64 </div>
  65
  66 <!-- *********************************************************************** -->
  67 <div class="doc_section">
  68   <a name="requirements">Keeping LLVM Portable</a>
  69 </div>
  70 <div class="doc_text">
  71   <p>In order to keep LLVM portable, LLVM developers should adhere to a set of
  72   portability rules associated with the System Library. Adherence to these rules
  73   should help the System Library achieve its goal of shielding LLVM from the
  74   variations in operating system interfaces and doing so efficiently.  The
  75   following sections define the rules needed to fulfill this objective.</p>
  76 </div>
  77
  78 <!-- ======================================================================= -->
  79 <div class="doc_subsection"><a name="headers">Don't Inlcude System Headers</a>
  80 </div>
  81 <div class="doc_text">
  82   <p>Except in <tt>lib/System</tt>, no LLVM source code should directly
  83   <tt>#include</tt> a system header. Care has been taken to remove all such
  84   <tt>#includes</tt> from LLVM while <tt>lib/System</tt> was being
  85   developed.  Specifically this means that header files like "unistd.h",
  86   "windows.h", "stdio.h", and "string.h" are forbidden to be included by LLVM
  87   source code outside the implementation of <tt>lib/System</tt>.</p>
  88   <p>To obtain system-dependent functionality, existing interfaces to the system
  89   found in <tt>include/llvm/System</tt> should be used. If an appropriate
  90   interface is not available, it should be added to <tt>include/llvm/System</tt>
  91   and implemented in <tt>lib/System</tt> for all supported platforms.</p>
  92 </div>
  93
  94 <!-- ======================================================================= -->
  95 <div class="doc_subsection"><a name="expose">Don't Expose System Headers</a>
  96 </div>
  97 <div class="doc_text">
  98   <p>The System Library must shield LLVM from <em>all</em> system headers. To
  99   obtain system level functionality, LLVM source must
 100   <tt>#include "llvm/System/Thing.h"</tt> and nothing else. This means that
 101   <tt>Thing.h</tt> cannot expose any system header files. This protects LLVM
 102   from accidentally using system specific functionality and only allows it
 103   via the <tt>lib/System</tt> interface.</p>
 104 </div>
 105
 106 <!-- ======================================================================= -->
 107 <div class="doc_subsection"><a name="c_headers">Use Standard C Headers</a></div>
 108 <div class="doc_text">
 109   <p>The <em>standard</em> C headers (the ones beginning with "c") are allowed
 110   to be exposed through the <tt>lib/System</tt> interface. These headers and
 111   the things they declare are considered to be platform agnostic. LLVM source
 112   files may include them directly or obtain their inclusion through
 113   <tt>lib/System</tt> interfaces.</p>
 114 </div>
 115
 116 <!-- ======================================================================= -->
 117 <div class="doc_subsection"><a name="cpp_headers">Use Standard C++ Headers</a>
 118 </div>
 119 <div class="doc_text">
 120   <p>The <em>standard</em> C++ headers from the standard C++ library and
 121   standard template library may be exposed through the <tt>lib/System</tt>
 122   interface. These headers and the things they declare are considered to be
 123   platform agnostic. LLVM source files may include them or obtain their
 124   inclusion through lib/System interfaces.</p>
 125 </div>
 126
 127 <!-- ======================================================================= -->
 128 <div class="doc_subsection"><a name="highlev">High Level Interface</a></div>
 129 <div class="doc_text">
 130   <p>The entry points specified in the interface of lib/System must be aimed at
 131   completing some reasonably high level task needed by LLVM. We do not want to
 132   simply wrap each operating system call. It would be preferable to wrap several
 133   operating system calls that are always used in conjunction with one another by
 134   LLVM.</p>
 135   <p>For example, consider what is needed to execute a program, wait for it to
 136   complete, and return its result code. On Unix, this involves the following
 137   operating system calls: <tt>getenv, fork, execve,</tt> and <tt>wait</tt>. The
 138   correct thing for lib/System to provide is a function, say
 139   <tt>ExecuteProgramAndWait</tt>, that implements the functionality completely.
 140   what we don't want is wrappers for the operating system calls involved.</p>
 141   <p>There must <em>not</em> be a one-to-one relationship between operating
 142   system calls and the System library's interface. Any such interface function
 143   will be suspicious.</p>
 144 </div>
 145
 146 <!-- ======================================================================= -->
 147 <div class="doc_subsection"><a name="nounused">No Unused Functionality</a></div>
 148 <div class="doc_text">
 149   <p>There must be no functionality specified in the interface of lib/System
 150   that isn't actually used by LLVM. We're not writing a general purpose
 151   operating system wrapper here, just enough to satisfy LLVM's needs. And, LLVM
 152   doesn't need much. This design goal aims to keep the lib/System interface
 153   small and understandable which should foster its actual use and adoption.</p>
 154 </div>
 155
 156 <!-- ======================================================================= -->
 157 <div class="doc_subsection"><a name="nodupl">No Duplicate Implementations</a>
 158 </div>
 159 <div class="doc_text">
 160   <p>The implementation of a function for a given platform must be written
 161   exactly once. This implies that it must be possible to apply a function's
 162   implementation to multiple operating systems if those operating systems can
 163   share the same implementation. This rule applies to the set of operating
 164   systems supported for a given class of operating system (e.g. Unix, Win32).
 165   </p>
 166 </div>
 167
 168 <!-- ======================================================================= -->
 169 <div class="doc_subsection"><a name="virtuals">No Virtual Methods</a></div>
 170 <div class="doc_text">
 171   <p>The System Library interfaces can be called quite frequently by LLVM. In
 172   order to make those calls as efficient as possible, we discourage the use of
 173   virtual methods. There is no need to use inheritance for implementation
 174   differences, it just adds complexity. The <tt>#include</tt> mechanism works
 175   just fine.</p>
 176 </div>
 177
 178 <!-- ======================================================================= -->
 179 <div class="doc_subsection"><a name="nofunc">No Exposed Functions</a></div>
 180 <div class="doc_text">
 181   <p>Any functions defined by system libraries (i.e. not defined by lib/System)
 182   must not be exposed through the lib/System interface, even if the header file
 183   for that function is not exposed. This prevents inadvertent use of system
 184   specific functionality.</p>
 185   <p>For example, the <tt>stat</tt> system call is notorious for having
 186   variations in the data it provides. <tt>lib/System</tt> must not declare
 187   <tt>stat</tt> nor allow it to be declared. Instead it should provide its own
 188   interface to discovering information about files and directories. Those
 189   interfaces may be implemented in terms of <tt>stat</tt> but that is strictly
 190   an implementation detail. The interface provided by the System Library must
 191   be implemented on all platforms (even those without <tt>stat</tt>).</p>
 192 </div>
 193
 194 <!-- ======================================================================= -->
 195 <div class="doc_subsection"><a name="nodata">No Exposed Data</a></div>
 196 <div class="doc_text">
 197   <p>Any data defined by system libraries (i.e. not defined by lib/System) must
 198   not be exposed through the lib/System interface, even if the header file for
 199   that function is not exposed. As with functions, this prevents inadvertent use
 200   of data that might not exist on all platforms.</p>
 201 </div>
 202
 203 <!-- ======================================================================= -->
 204 <div class="doc_subsection"><a name="softerrors">Minimize Soft Errors</a></div>
 205 <div class="doc_text">
 206   <p>Operating system interfaces will generally provide error results for every
 207   little thing that could go wrong. In almost all cases, you can divide these
 208   error results into two groups: normal/good/soft and abnormal/bad/hard. That
 209   is, some of the errors are simply information like "file not found",
 210   "insufficient privileges", etc. while other errors are much harder like
 211   "out of space", "bad disk sector", or "system call interrupted". We'll call
 212   the first group "<i>soft</i>" errors and the second group "<i>hard</i>"
 213   errors.<p>
 214   <p>lib/System must always attempt to minimize soft errors and always just
 215   throw a std::string on hard errors. This is a design requirement because the
 216   minimization of soft errors can affect the granularity and the nature of the
 217   interface. In general, if you find that you're wanting to throw soft errors,
 218   you must review the granularity of the interface because it is likely you're
 219   trying to implement something that is too low level. The rule of thumb is to
 220   provide interface functions that <em>can't</em> fail, except when faced with
 221   hard errors.</p>
 222   <p>For a trivial example, suppose we wanted to add an "OpenFileForWriting"
 223   function. For many operating systems, if the file doesn't exist, attempting
 224   to open the file will produce an error.  However, lib/System should not
 225   simply throw that error if it occurs because its a soft error. The problem
 226   is that the interface function, OpenFileForWriting is too low level. It should
 227   be OpenOrCreateFileForWriting. In the case of the soft "doesn't exist" error,
 228   this function would just create it and then open it for writing.</p>
 229   <p>This design principle needs to be maintained in lib/System because it
 230   avoids the propagation of soft error handling throughout the rest of LLVM.
 231   Hard errors will generally just cause a termination for an LLVM tool so don't
 232   be bashful about throwing them.</p>
 233   <p>Rules of thumb:</p>
 234   <ol>
 235     <li>Don't throw soft errors, only hard errors.</li>
 236     <li>If you're tempted to throw a soft error, re-think the interface.</li>
 237     <li>Handle internally the most common normal/good/soft error conditions
 238     so the rest of LLVM doesn't have to.</li>
 239   </ol>
 240 </div>
 241
 242 <!-- ======================================================================= -->
 243 <div class="doc_subsection"><a name="throw">Throw Only std::string</a></div>
 244 <div class="doc_text">
 245   <p>If an error occurs that lib/System cannot handle, the only action taken by
 246   lib/System is to throw an instance of std:string. The contents of the string
 247   must explain both what happened and the context in which it happened. The
 248   format of the string should be a (possibly empty) list of contexts each
 249   terminated with a : and a space, followed by the error message, optionally
 250   followed by a reason, and optionally followed by a suggestion.</p>
 251   <p>For example, failure to open a file named "foo" could result in a message
 252   like:</p>
 253   <ul><li>foo: Unable to open file because it doesn't exist."</li></ul>
 254   <p>The "foo:" part is the context. The "Unable to open file" part is the error
 255   message. The "because it doesn't exist." part is the reason. This message has
 256   no suggestion. Where possible, the implementation of lib/System should use
 257   operating system specific facilities for converting the error code returned by
 258   a system call into an error message. This will help to make the error message
 259   more familiar to users of that type of operating system.</p>
 260   <p>Note that this requirement precludes the throwing of any other exceptions.
 261   For example, various C++ standard library functions can cause exceptions to be
 262   thrown (e.g. out of memory situation). In all cases, if there is a possibility
 263   that non-string exceptions could be thrown, the lib/System library must ensure
 264   that the exceptions are translated to std::string form.</p>
 265 </div>
 266
 267 <!-- ======================================================================= -->
 268 <div class="doc_subsection"><a name="throw_spec">No throw Specifications</a>
 269 </div>
 270 <div class="doc_text">
 271   <p>None of the lib/System interface functions may be declared with C++
 272   <tt>throw()</tt> specifications on them. This requirement makes sure that the
 273   compiler does not insert additional exception handling code into the interface
 274   functions. This is a performance consideration: lib/System functions are at
 275   the bottom of many call chains and as such can be frequently called. We
 276   need them to be as efficient as possible.</p>
 277 </div>
 278
 279 <!-- ======================================================================= -->
 280 <div class="doc_subsection"><a name="organization">Code Organization</a></div>
 281 <div class="doc_text">
 282   <p>Implementations of the System Library interface are separated by their
 283   general class of operating system. Currently only Unix and Win32 classes are
 284   defined but more could be added for other operating system classifications.
 285   To distinguish which implementation to compile, the code in lib/System uses
 286   the LLVM_ON_UNIX and LLVM_ON_WIN32 #defines provided via configure through the
 287   llvm/Config/config.h file. Each source file in lib/System, after implementing
 288   the generic (operating system independent) functionality needs to include the
 289   correct implementation using a set of <tt>#if defined(LLVM_ON_XYZ)</tt>
 290   directives. For example, if we had lib/System/File.cpp, we'd expect to see in
 291   that file:</p>
 292   <pre><tt>
 293   #if defined(LLVM_ON_UNIX)
 294   #include "Unix/File.cpp"
 295   #endif
 296   #if defined(LLVM_ON_WIN32)
 297   #include "Win32/File.cpp"
 298   #endif
 299   </tt></pre>
 300   <p>The implementation in lib/System/Unix/File.cpp should handle all Unix
 301   variants. The implementation in lib/System/Win32/File.cpp should handle all
 302   Win32 variants.  What this does is quickly differentiate the basic class of
 303   operating system that will provide the implementation. The specific details
 304   for a given platform must still be determined through the use of
 305   <tt>#ifdef</tt>.</p>
 306 </div>
 307
 308 <!-- ======================================================================= -->
 309 <div class="doc_subsection"><a name="semantics">Consistent Semantics</a></div>
 310 <div class="doc_text">
 311   <p>The implementation of a lib/System interface can vary drastically between
 312   platforms. That's okay as long as the end result of the interface function
 313   is the same. For example, a function to create a directory is pretty straight
 314   forward on all operating system. System V IPC on the other hand isn't even
 315   supported on all platforms. Instead of "supporting" System V IPC, lib/System
 316   should provide an interface to the basic concept of inter-process
 317   communications. The implementations might use System V IPC if that was
 318   available or named pipes, or whatever gets the job done effectively for a
 319   given operating system.  In all cases, the interface and the implementation
 320   must be semantically consistent. </p>
 321 </div>
 322
 323 <!-- ======================================================================= -->
 324 <div class="doc_subsection"><a name="bug">Bug 351</a></div>
 325 <div class="doc_text">
 326   <p>See <a href="http://llvm.org/PR351">bug 351</a>
 327   for further details on the progress of this work</p>
 328 </div>
 329
 330 <!-- *********************************************************************** -->
 331
 332 <hr>
 333 <address>
 334   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
 335   src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
 336   <a href="http://validator.w3.org/check/referer"><img
 337   src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
 338
 339   <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br>
 340   <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
 341   Last modified: $Date$
 342 </address>
 343 </body>
 344 </html>