| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 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 |  | 
| Misha Brukman | ebbbf25 | 2004-07-21 18:02:43 +0000 | [diff] [blame] | 10 | <div class="doc_title">System Library</div> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 11 |  | 
| Misha Brukman | ebbbf25 | 2004-07-21 18:02:43 +0000 | [diff] [blame] | 12 | <div class="doc_warning"> | 
 | 13 |   <p>Warning: This document is a work in progress.</p> | 
 | 14 | </div> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 15 |  | 
 | 16 | <ul> | 
 | 17 |   <li><a href="#abstract">Abstract</a></li> | 
 | 18 |   <li><a href="#requirements">System Library Requirements</a> | 
 | 19 |   <ol> | 
 | 20 |     <li><a href="#headers">Hide System Header Files</a></li> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 21 |     <li><a href="#c_headers">Allow Standard C Header Files</a></li> | 
 | 22 |     <li><a href="#cpp_headers">Allow Standard C++ Header Files</a></li> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 23 |     <li><a href="#nofunc">No Exposed Functions</a></li> | 
 | 24 |     <li><a href="#nodata">No Exposed Data</a></li> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 25 |     <li><a href="#throw">Throw Only std::string</a></li> | 
 | 26 |     <li><a href="#throw_spec">No throw() Specifications</a></li> | 
 | 27 |     <li><a href="#nodupl">No Duplicate Impementations</a></li> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 28 |   </ol></li> | 
 | 29 |   <li><a href="#design">System Library Design</a> | 
 | 30 |   <ol> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 31 |     <li><a href="#nounused">No Unused Functionality</a></li> | 
 | 32 |     <li><a href="#highlev">High-Level Interface</a></li> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 33 |     <li><a href="#opaque">Use Opaque Classes</a></li> | 
 | 34 |     <li><a href="#common">Common Implementations</a></li> | 
 | 35 |     <li><a href="#multi_imps">Multiple Implementations</a></li> | 
 | 36 |     <li><a href="#lowlevel">Use Low Level Interfaces</a></li> | 
 | 37 |     <li><a href="#memalloc">No Memory Allocation</a></li> | 
 | 38 |     <li><a href="#virtuals">No Virtual Methods</a></li> | 
 | 39 |   </ol></li> | 
 | 40 |   <li><a href="#detail">System Library Details</a> | 
 | 41 |   <ol> | 
 | 42 |     <li><a href="#bug">Tracking Bugzilla Bug: 351</a></li> | 
 | 43 |     <li><a href="#refimpl">Reference Implementatation</a></li> | 
 | 44 |   </ol></li> | 
 | 45 | </ul> | 
 | 46 |  | 
 | 47 | <div class="doc_author"> | 
| Misha Brukman | 1721784 | 2004-07-21 18:04:27 +0000 | [diff] [blame] | 48 |   <p>Written by <a href="rspencer@x10sys.com">Reid Spencer</a></p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 49 | </div> | 
 | 50 |  | 
 | 51 |  | 
 | 52 | <!-- *********************************************************************** --> | 
 | 53 | <div class="doc_section"><a name="abstract">Abstract</a></div> | 
 | 54 | <div class="doc_text"> | 
 | 55 |   <p>This document describes the requirements, design, and implementation  | 
 | 56 |   details of LLVM's System Library. The library is composed of the header files | 
 | 57 |   in <tt>llvm/include/llvm/System</tt> and the source files in  | 
 | 58 |   <tt>llvm/lib/System</tt>. The goal of this library is to completely shield  | 
 | 59 |   LLVM from the variations in operating system interfaces. By centralizing  | 
 | 60 |   LLVM's use of operating system interfaces, we make it possible for the LLVM | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 61 |   tool chain and runtime libraries to be more easily ported to new platforms | 
 | 62 |   since (theoretically) only <tt>llvm/lib/System</tt> needs to be ported.  This | 
 | 63 |   library also unclutters the rest of LLVM from #ifdef use and special | 
 | 64 |   cases for specific operating systems. Such uses are replaced with simple calls | 
 | 65 |   to the interfaces provided in <tt>llvm/include/llvm/System</tt>.</p> Note that | 
 | 66 |   lib/System is not intended to be a complete operating system wrapper (such as | 
 | 67 |   the Adaptive Communications Environment (ACE) or Apache Portable Runtime | 
 | 68 |   (APR)), but only to provide the functionality necessary to support LLVM. | 
 | 69 |   <p>The System Library was written by Reid Spencer who formulated the | 
 | 70 |   design based on similar original work as part of the eXtensible Programming  | 
 | 71 |   System (XPS).</p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 72 | </div> | 
 | 73 |  | 
 | 74 | <!-- *********************************************************************** --> | 
 | 75 | <div class="doc_section"> | 
 | 76 |   <a name="requirements">System Library Requirements</a> | 
 | 77 | </div> | 
 | 78 | <div class="doc_text"> | 
 | 79 |   <p>The System library's requirements are aimed at shielding LLVM from the | 
 | 80 |   variations in operating system interfaces. The following sections define the | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 81 |   requirements needed to fulfill this objective. Of necessity, these requirements  | 
 | 82 |   must be strictly followed in order to ensure the library's goal is reached.</p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 83 | </div> | 
 | 84 |  | 
 | 85 | <!-- ======================================================================= --> | 
 | 86 | <div class="doc_subsection"><a name="headers">Hide System Header Files</a></div> | 
 | 87 | <div class="doc_text"> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 88 |   <p>The library must sheild LLVM from <em>all</em> system libraries. To obtain | 
 | 89 |   system level functionality, LLVM must <tt>#include "llvm/System/Thing.h"</tt> | 
 | 90 |   and nothing else. This means that <tt>Thing.h</tt> cannot expose any system | 
 | 91 |   header files. This protects LLVM from accidentally using system specific | 
 | 92 |   functionality except through the lib/System interface.  Specifically this  | 
 | 93 |   means that header files like "unistd.h", "windows.h", "stdio.h", and  | 
 | 94 |   "string.h" are verbotten outside the implementation of lib/System. | 
 | 95 |   </p> | 
 | 96 | </div> | 
 | 97 |  | 
 | 98 | <!-- ======================================================================= --> | 
 | 99 | <div class="doc_subsection"><a name="c_headers">Allow Standard C Headers</a> | 
 | 100 | </div> | 
 | 101 | <div class="doc_text"> | 
 | 102 |   <p>The <em>standard</em> C headers (the ones beginning with "c") are allowed | 
 | 103 |   to be exposed through the lib/System interface. These headers and the things | 
 | 104 |   they declare are considered to be platform agnostic. LLVM source files may | 
 | 105 |   include them or obtain their inclusion through lib/System interfaces.</p> | 
 | 106 | </div> | 
 | 107 |  | 
 | 108 | <!-- ======================================================================= --> | 
 | 109 | <div class="doc_subsection"><a name="cpp_headers">Allow Standard C++ Headers</a> | 
 | 110 | </div> | 
 | 111 | <div class="doc_text"> | 
 | 112 |   <p>The <em>standard</em> C++ headers from the standard C++ library and | 
 | 113 |   standard template library are allowed to be exposed through the lib/System | 
 | 114 |   interface. These headers and the things they declare are considered to be | 
 | 115 |   platform agnostic. LLVM source files may include them or obtain their | 
 | 116 |   inclusion through lib/System interfaces.</p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 117 | </div> | 
 | 118 |  | 
 | 119 | <!-- ======================================================================= --> | 
 | 120 | <div class="doc_subsection"><a name="nofunc">No Exposed Functions</a></div> | 
 | 121 | <div class="doc_text"> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 122 |   <p>Any functions defined by system libraries (i.e. not defined by lib/System)  | 
 | 123 |   must not be exposed through the lib/System interface, even if the header file  | 
 | 124 |   for that function is not exposed. This prevents inadvertent use of system | 
 | 125 |   specific functionality.</p> | 
 | 126 |   <p>For example, the <tt>stat</tt> system call is notorious for having | 
 | 127 |   variations in the data it provides. lib/System must not declare <tt>stat</tt> | 
 | 128 |   nor allow it to be declared. Instead it should provide its own interface to | 
 | 129 |   discovering information about files and directories. Those interfaces may be | 
 | 130 |   implemented in terms of <tt>stat</tt> but that is strictly an implementation | 
 | 131 |   detail.</p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 132 | </div> | 
 | 133 |  | 
 | 134 | <!-- ======================================================================= --> | 
 | 135 | <div class="doc_subsection"><a name="nodata">No Exposed Data</a></div> | 
 | 136 | <div class="doc_text"> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 137 |   <p>Any data defined by system libraries (i.e. not defined by lib/System) must | 
 | 138 |   not be exposed through the lib/System interface, even if the header file for | 
 | 139 |   that function is not exposed. As with functions, this prevents inadvertent use | 
 | 140 |   of data that might not exist on all platforms.</p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 141 | </div> | 
 | 142 |  | 
 | 143 | <!-- ======================================================================= --> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 144 | <div class="doc_subsection"><a name="throw">Throw Only std::string</a></div> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 145 | <div class="doc_text"> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 146 |   <p>If an error occurs that lib/System cannot handle, the only action taken by | 
 | 147 |   lib/System is to throw an instance of std:string. The contents of the string | 
 | 148 |   must explain both what happened and the context in which it happened. The | 
 | 149 |   format of the string should be a (possibly empty) list of contexts each  | 
 | 150 |   terminated with a : and a space, followed by the error message, optionally | 
 | 151 |   followed by a reason, and optionally followed by a suggestion.</p> | 
 | 152 |   <p>For example, failure to open a file named "foo" could result in a message | 
 | 153 |   like:</p> | 
 | 154 |   <ul><li>foo: Unable to open file because it doesn't exist."</li></ul> | 
 | 155 |   <p>The "foo:" part is the context. The "Unable to open file" part is the error | 
 | 156 |   message. The "because it doesn't exist." part is the reason. This message has | 
 | 157 |   no suggestion. Where possible, the imlementation of lib/System should use | 
 | 158 |   operating system specific facilities for converting the error code returned by | 
 | 159 |   a system call into an error message. This will help to make the error message | 
 | 160 |   more familiar to users of that type of operating system.</p> | 
 | 161 |   <p>Note that this requirement precludes the throwing of any other exceptions. | 
 | 162 |   For example, various C++ standard library functions can cause exceptions to be | 
 | 163 |   thrown (e.g. out of memory situation). In all cases, if there is a possibility | 
 | 164 |   that non-string exceptions could be thrown, the lib/System library must ensure | 
 | 165 |   that the exceptions are translated to std::string form.</p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 166 | </div> | 
 | 167 |  | 
 | 168 | <!-- ======================================================================= --> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 169 | <div class="doc_subsection"><a name="throw_spec">No throw Specifications</a> | 
 | 170 | </div> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 171 | <div class="doc_text"> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 172 |   <p>None of the lib/System interface functions may be declared with C++  | 
 | 173 |   <tt>throw()</tt> specifications on them. This requirement makes sure that the | 
 | 174 |   compler does not insert addtional exception handling code into the interface | 
 | 175 |   functions. This is a performance consideration: lib/System functions are at | 
 | 176 |   the bottom of the many call chains and as such can be frequently called. We | 
 | 177 |   need them to be as efficient as possible.</p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 178 | </div> | 
 | 179 |  | 
 | 180 | <!-- ======================================================================= --> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 181 | <div class="doc_subsection"><a name="nodupl">No Duplicate Implementations</a> | 
 | 182 | </div> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 183 | <div class="doc_text"> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 184 |   <p>The implementation of a function for a given platform must be written | 
 | 185 |   exactly once. This implies that it must be possible to apply a function's  | 
 | 186 |   implementation to multiple operating systems if those operating systems can | 
 | 187 |   share the same implementation.</p> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 188 | </div> | 
 | 189 |  | 
 | 190 | <!-- *********************************************************************** --> | 
 | 191 | <div class="doc_section"><a name="design">System Library Design</a></div> | 
 | 192 | <div class="doc_text"> | 
 | 193 |   <p>In order to fulfill the requirements of the system library, strict design | 
 | 194 |   objectives must be maintained in the library as it evolves.  The goal here  | 
 | 195 |   is to provide interfaces to operating system concepts (files, memory maps,  | 
 | 196 |   sockets, signals, locking, etc) efficiently and in such a way that the  | 
 | 197 |   remainder of LLVM is completely operating system agnostic.</p> | 
 | 198 | </div> | 
 | 199 |  | 
 | 200 | <!-- ======================================================================= --> | 
| Reid Spencer | 7acb866 | 2004-08-27 02:08:04 +0000 | [diff] [blame] | 201 | <div class="doc_subsection"><a name="nounused">No Unused Functionality</a></div> | 
 | 202 | <div class="doc_text"> | 
 | 203 |   <p>There must be no functionality specified in the interface of lib/System  | 
 | 204 |   that isn't actually used by LLVM. We're not writing a general purpose | 
 | 205 |   operating system wrapper here, just enough to satisfy LLVM's needs. And, LLVM | 
 | 206 |   doesn't need much. This design goal aims to keep the lib/System interface | 
 | 207 |   small and understandable which should foster its actual use and adoption.</p> | 
 | 208 | </div> | 
 | 209 |  | 
 | 210 | <!-- ======================================================================= --> | 
 | 211 | <div class="doc_subsection"><a name="highlev">High Level Interface</a></div> | 
 | 212 | <div class="doc_text"> | 
 | 213 |   <p>The entry points specified in the interface of lib/System must be aimed at  | 
 | 214 |   completing some reasonably high level task needed by LLVM. We do not want to | 
 | 215 |   simply wrap each operating system call. It would be preferable to wrap several | 
 | 216 |   operating system calls that are always used in conjunction with one another by | 
 | 217 |   LLVM.</p> | 
 | 218 |   <p>For example, consider what is needed to execute a program, wait for it to | 
 | 219 |   complete, and return its result code. On Unix, this involves the following | 
 | 220 |   operating system calls: <tt>getenv, fork, execve,</tt> and <tt>wait</tt>. The | 
 | 221 |   correct thing for lib/System to provide is a function, say | 
 | 222 |   <tt>ExecuteProgramAndWait</tt>, that implements the functionality completely. | 
 | 223 |   what we don't want is wrappers for the operating system calls involved.</p> | 
 | 224 |   <p>There must <em>not</em> be a one-to-one relationship between operating | 
 | 225 |   system calls and the System library's interface. Any such interface function | 
 | 226 |   will be suspicious.</p> | 
 | 227 | </div> | 
 | 228 |  | 
 | 229 | <!-- ======================================================================= --> | 
 | 230 | <div class="doc_subsection"><a name="highlev">Minimize Soft Errors</a></div> | 
 | 231 | <div class="doc_text"> | 
 | 232 |   <p>Operating system interfaces will generally provide errors results for every | 
 | 233 |   little thing that could go wrong. In almost all cases, you can divide these | 
 | 234 |   error results into two groups: normal/good/soft and abnormal/bad/hard. That | 
 | 235 |   is, some of the errors are simply information like "file not found",  | 
 | 236 |   "insufficient privileges", etc. while other errors are much harder like | 
 | 237 |   "out of space", "bad disk sector", or "system call interrupted". Well call the | 
 | 238 |   first group "soft" errors and the second group "hard" errors.<p> | 
 | 239 |   <p>lib/System must always attempt to minimize soft errors and always just | 
 | 240 |   throw a std::string on hard errors. This is a design requirement because the | 
 | 241 |   minimization of soft errors can affect the granularity and the nature of the | 
 | 242 |   interface. In general, if you find that you're wanting to throw soft errors, | 
 | 243 |   you must review the granularity of the interface because it is likely you're | 
 | 244 |   trying to implement something that is too low level. The rule of thumb is to | 
 | 245 |   provide interface functions that "can't" fail, except when faced with hard | 
 | 246 |   errors.</p> | 
 | 247 |   <p>For a trivial example, suppose we wanted to add an "OpenFileForWriting"  | 
 | 248 |   function. For many operating systems, if the file doesn't exist, attempting  | 
 | 249 |   to open the file will produce an error.  However, lib/System should not | 
 | 250 |   simply throw that error if it occurs because its a soft error. The problem | 
 | 251 |   is that the interface function, OpenFileForWriting is too low level. It should | 
 | 252 |   be OpenOrCreateFileForWriting. In the case of the soft "doesn't exist" error,  | 
 | 253 |   this function would just create it and then open it for writing.</p> | 
 | 254 |   <p>This design principle needs to be maintained in lib/System because it | 
 | 255 |   avoids the propagation of soft error handling throughout the rest of LLVM. | 
 | 256 |   Hard errors will generally just cause a termination for an LLVM tool so don't | 
 | 257 |   be bashful about throwing them.</p> | 
 | 258 |   <p>Rules of thumb:</p> | 
 | 259 |   <ol> | 
 | 260 |     <li>Don't throw soft errors, only hard errors.</li> | 
 | 261 |     <li>If you're tempted to throw a soft error, re-think the interface.</li> | 
 | 262 |     <li>Handle internally the most common normal/good/soft error conditions | 
 | 263 |     so the rest of LLVM doesn't have to.</li> | 
 | 264 |   </ol> | 
 | 265 |   | 
 | 266 | <pre><tt> | 
 | 267 | Notes: | 
 | 268 | 10. The implementation of a lib/System interface can vary drastically between | 
 | 269 |     platforms. That's okay as long as the end result of the interface function is | 
 | 270 |     the same. For example, a function to create a directory is pretty straight | 
 | 271 |     forward on all operating system. System V IPC on the other hand isn't even | 
 | 272 |     supported on all platforms. Instead of "supporting" System V IPC, lib/System | 
 | 273 |     should provide an interface to the basic concept of inter-process  | 
 | 274 |     communications. The implementations might use System V IPC if that was | 
 | 275 |     available or named pipes, or whatever gets the job done effectively for a | 
 | 276 |     given operating system. | 
 | 277 |  | 
 | 278 | 11. Implementations are separated first by the general class of operating system | 
 | 279 |     as provided by the configure script's $build variable. This variable is used | 
 | 280 |     to create a link from $BUILD_OBJ_ROOT/lib/System/platform to a directory in | 
 | 281 |     $BUILD_SRC_ROOT/lib/System directory with the same name as the $build | 
 | 282 |     variable. This provides a retargetable include mechanism. By using the link's | 
 | 283 |     name (platform) we can actually include the operating specific | 
 | 284 |     implementation. For example, support $build is "Darwin" for MacOS X. If we | 
 | 285 |     place: | 
 | 286 |       #include "platform/File.cpp" | 
 | 287 |     into a a file in lib/System, it will actually include | 
 | 288 |     lib/System/Darwin/File.cpp. What this does is quickly differentiate the basic | 
 | 289 |     class of operating system that will provide the implementation. | 
 | 290 |   | 
 | 291 | 12. Implementation files in lib/System need may only do two things: (1) define  | 
 | 292 |     functions and data that is *TRULY* generic (completely platform agnostic) and | 
 | 293 |     (2) #include the platform specific implementation with: | 
 | 294 |   | 
 | 295 |        #include "platform/Impl.cpp" | 
 | 296 |   | 
 | 297 |     where Impl is the name of the implementation files. | 
 | 298 |   | 
 | 299 | 13. Platform specific implementation files (platform/Impl.cpp) may only #include | 
 | 300 |     other Impl.cpp files found in directories under lib/System. The order of | 
 | 301 |     inclusion is very important (from most generic to most specific) so that we | 
 | 302 |     don't inadvertently place an implementation in the wrong place. For example, | 
 | 303 |     consider a fictitious implementation file named DoIt.cpp. Here's how the | 
 | 304 |     #includes should work for a Linux platform | 
 | 305 |   | 
 | 306 |     lib/System/DoIt.cpp | 
 | 307 |       #include "platform/DoIt.cpp"        // platform specific impl. of Doit | 
 | 308 |       DoIt | 
 | 309 |   | 
 | 310 |     lib/System/Linux/DoIt.cpp             // impl that works on all Linux  | 
 | 311 |       #include "../Unix/DoIt.cpp"         // generic Unix impl. of DoIt | 
 | 312 |       #include "../Unix/SUS/DoIt.cpp      // SUS specific impl. of DoIt | 
 | 313 |       #include "../Unix/SUS/v3/DoIt.cpp   // SUSv3 specific impl. of DoIt | 
 | 314 |   | 
 | 315 |     Note that the #includes in lib/System/Linux/DoIt.cpp are all optional but | 
 | 316 |     should be used where the implementation of some functionality can be shared | 
 | 317 |     across some set of Unix variants. We don't want to duplicate code across | 
 | 318 |     variants if their implementation could be shared. | 
 | 319 | </tt></pre> | 
 | 320 | </div> | 
 | 321 |  | 
 | 322 | <!-- ======================================================================= --> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 323 | <div class="doc_subsection"><a name="opaque">Use Opaque Classes</a></div> | 
 | 324 | <div class="doc_text"> | 
 | 325 |   <p>no public data</p> | 
 | 326 |   <p>onlyprimitive typed private/protected data</p> | 
 | 327 |   <p>data size is "right" for platform, not max of all platforms</p> | 
 | 328 |   <p>each class corresponds to O/S concept</p> | 
 | 329 | </div> | 
 | 330 |  | 
 | 331 | <!-- ======================================================================= --> | 
 | 332 | <div class="doc_subsection"><a name="common">Common Implementations</a></div> | 
 | 333 | <div class="doc_text"> | 
 | 334 |   <p>To be written.</p> | 
 | 335 | </div> | 
 | 336 |  | 
 | 337 | <!-- ======================================================================= --> | 
 | 338 | <div class="doc_subsection"> | 
 | 339 |   <a name="multi_imps">Multiple Implementations</a> | 
 | 340 | </div> | 
 | 341 | <div class="doc_text"> | 
 | 342 |   <p>To be written.</p> | 
 | 343 | </div> | 
 | 344 |  | 
 | 345 | <!-- ======================================================================= --> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 346 | <div class="doc_subsection"><a name="memalloc">No Memory Allocation</a></div> | 
 | 347 | <div class="doc_text"> | 
 | 348 |   <p>To be written.</p> | 
 | 349 | </div> | 
 | 350 |  | 
 | 351 | <!-- ======================================================================= --> | 
 | 352 | <div class="doc_subsection"><a name="virtuals">No Virtual Methods</a></div> | 
 | 353 | <div class="doc_text"> | 
 | 354 |   <p>To be written.</p> | 
 | 355 | </div> | 
 | 356 |  | 
 | 357 | <!-- *********************************************************************** --> | 
 | 358 | <div class="doc_section"><a name="detail">System Library Details</a></div> | 
 | 359 | <div class="doc_text"> | 
 | 360 |   <p>To be written.</p> | 
 | 361 | </div> | 
 | 362 |  | 
 | 363 | <!-- ======================================================================= --> | 
| Chris Lattner | 05540a7 | 2004-07-17 18:50:19 +0000 | [diff] [blame] | 364 | <div class="doc_subsection"><a name="bug">Bug 351</a></div> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 365 | <div class="doc_text"> | 
| Chris Lattner | 05540a7 | 2004-07-17 18:50:19 +0000 | [diff] [blame] | 366 |   <p>See <a href="http://llvm.cs.uiuc.edu/PR351">bug 351</a> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 367 |   for further details on the progress of this work</p> | 
 | 368 | </div> | 
 | 369 |  | 
 | 370 | <!-- ======================================================================= --> | 
| Reid Spencer | 9518920 | 2004-08-26 18:52:52 +0000 | [diff] [blame] | 371 | <div class="doc_subsection"><a name="bug">Rationale For #include Hierarchy</a> | 
 | 372 | </div> | 
 | 373 | <div class="doc_text"> | 
 | 374 |   <p>In order to provide different implementations of the lib/System interface | 
 | 375 |   for different platforms, it is necessary for the library to "sense" which | 
 | 376 |   operating system is being compiled for and conditionally compile only the | 
 | 377 |   applicabe parts of the library. While several operating system wrapper | 
 | 378 |   libraries (e.g. APR, ACE) choose to use #ifdef preprocessor statements in | 
 | 379 |   combination with autoconf variable (HAVE_* family), lib/System chooses an | 
 | 380 |   alternate strategy. <p> | 
 | 381 |   <p>To put it succinctly, the lib/System strategy has traded "#ifdef hell" for  | 
 | 382 |   "#include hell". That is, a given implementation file defines one or more | 
 | 383 |   functions for a particular operating system variant. The functions defined in | 
 | 384 |   that file have no #ifdef's to disambiguate the platform since the file is only | 
 | 385 |   compiled on one kind of platform. While this leads to the same function being | 
 | 386 |   imlemented differently in different files, it is our contention that this | 
 | 387 |   leads to better maintenance and easier portability.</p> | 
 | 388 |   <p>For example, consider a function having different implementations on a | 
 | 389 |   variety of platforms. Many wrapper libraries choose to deal with the different | 
 | 390 |   implementations by using #ifdef, like this:</p> | 
 | 391 |   <pre><tt> | 
 | 392 |       void SomeFunction(void) { | 
 | 393 |       #if defined __LINUX | 
 | 394 |         // .. Linux implementation | 
 | 395 |       #elif defined __WIN32 | 
 | 396 |         // .. Win32 implementation | 
 | 397 |       #elif defined __SunOS | 
 | 398 |         // .. SunOS implementation | 
 | 399 |       #else | 
 | 400 |       #warning "Don't know how to implement SomeFunction on this platform" | 
 | 401 |       #endif | 
 | 402 |       } | 
 | 403 |   </tt></pre> | 
 | 404 |   <p>The problem with this is that its very messy to read, especially as the | 
 | 405 |   number of operating systems and their variants grow. The above example is | 
 | 406 |   actually tame compared to what can happen when the implementation depends on | 
 | 407 |   specific flavors and versions of the operating system. In that case you end up | 
 | 408 |   with multiple levels of nested #if statements. This is what we mean by "#ifdef | 
 | 409 |   hell".</p> | 
 | 410 |   <p>To avoid the situation above, we've choosen to locate all functions for a | 
 | 411 |   given implementation file for a specific operating system into one place. This | 
 | 412 |   has the following advantages:<p> | 
 | 413 |   <ul> | 
 | 414 |     <li>No "#ifdef hell"</li> | 
 | 415 |     <li>When porting, the strategy is quite straight forward: copy the | 
 | 416 |     implementation file from a similar operating system to a new directory and | 
 | 417 |     re-implement them.<li> | 
 | 418 |     <li>Correctness is helped during porting because the new operating system's | 
 | 419 |     implementation is wholly contained in a separate directory. There's no | 
 | 420 |     chance to make an error in the #if statements and affect some other | 
 | 421 |     operating system's implementation.</li> | 
 | 422 |   </ul> | 
 | 423 |   <p>So, given that we have decided to use #include instead of #if to provide | 
 | 424 |   platform specific implementations, there are actually three ways we can go | 
 | 425 |   about doing this. None of them are perfect, but we believe we've chosen the | 
 | 426 |   lesser of the three evils. Given that there is a variable named $OS which | 
 | 427 |   names the platform for which we must build, here's a summary of the three  | 
 | 428 |   approaches we could use to determine the correct directory:</p> | 
 | 429 |   <ol> | 
 | 430 |     <li>Provide the compiler with a -I$(OS) on the command line. This could be | 
 | 431 |     provided in only the lib/System makefile.</li> | 
 | 432 |     <li>Use autoconf to transform #include statements in the implementation | 
 | 433 |     files by using substitutions of @OS@. For example, if we had a file, | 
 | 434 |     File.cpp.in, that contained "#include <@OS@/File.cpp>" this would get | 
 | 435 |     transformed to "#include <actual/File.cpp>" where "actual" is the | 
 | 436 |     actual name of the operating system</li> | 
 | 437 |     <li>Create a link from $OBJ_DIR/platform to $SRC_DIR/$OS. This allows us to | 
 | 438 |     use a generic directory name to get the correct platform, as in #include | 
 | 439 |     <platform/File.cpp></li> | 
 | 440 |   </ol> | 
 | 441 |   <p>Let's look at the pitfalls of each approach.</p> | 
 | 442 |   <p>In approach #1, we end up with some confusion as to what gets included. | 
 | 443 |   Suppose we have lib/System/File.cpp that includes just File.cpp to get the | 
 | 444 |   platform specific part of the implementation. In this case, the include | 
 | 445 |   directive with the <> syntax will include the right file but the include | 
 | 446 |   directive with the "" syntax will recursively include the same file, | 
 | 447 |   lib/System/File.cpp. In the case of #include <File.cpp>, the -I options | 
 | 448 |   to the compiler are searched first so it works. But in the #include "File.cpp" | 
 | 449 |   case, the current directory is searched first. Furthermore, in both cases, | 
 | 450 |   neither include directive documents which File.cpp is getting included.</p> | 
 | 451 |   <p>In approach #2, we have the problem of needing to reconfigure repeatedly. | 
 | 452 |   Developer's generally hate that and we don't want lib/System to be a thorn in | 
 | 453 |   everyone's side because it will constantly need updating as operating systems | 
 | 454 |   change and as new operating systems are added. The problem occurs when a new | 
 | 455 |   implementation file is added to the library. First of all, you have to add a | 
 | 456 |   file with the .in suffix, then you have to add that file name to the list of | 
 | 457 |   configurable files in the autoconf/configure.ac file, then you have to run | 
 | 458 |   AutoRegen.sh to rebuild the configure script, then you have to run the | 
 | 459 |   configure script. This is deemed to be a pretty large hassle.</p> | 
 | 460 |   <p>In approach #3, we have the problem that not all platforms support links. | 
 | 461 |   Fortunately the autoconf macro used to create the link can compensate for | 
 | 462 |   this. If a link can't be made, the configure script will copy the correct | 
 | 463 |   directory from $BUILD_SRC_DIR to $BUILD_OBJ_DIR under the new name. The only | 
 | 464 |   problem with this is that if a copy is made, the copy doesn't get updated if | 
 | 465 |   the programmer adds or modifies files in the $BUILD_SRC_DIR. A reconfigure or | 
 | 466 |   manual copying is needed to get things to compile.<p> | 
 | 467 |   <p>The approach we have taken in lib/System is #3. Here's why:<p> | 
 | 468 |   <ul> | 
 | 469 |     <li>Approach #1 is rejected because it doesn't document what's actually | 
 | 470 |     getting included and the potential for mistakes with alternate include | 
 | 471 |     directive forms is high.</li> | 
 | 472 |     <li>Approach #2 are both viable and only really impact development when new | 
 | 473 |     files are added to the library.</li> | 
 | 474 |     <li>However, approach #2 impacts every new file on every platform all the | 
 | 475 |     time. With approach #3, only those platforms not supporting links will be | 
 | 476 |     affected. The number of platforms not supporting links is very small and | 
 | 477 |     they are generally archaic.</li> | 
 | 478 |     <li>Given the above, approach #3 seems to have the least impact.</li> | 
 | 479 |   </ul> | 
 | 480 | </div> | 
 | 481 |  | 
 | 482 | <!-- ======================================================================= --> | 
| Reid Spencer | 4f6015a | 2004-07-17 10:04:49 +0000 | [diff] [blame] | 483 | <div class="doc_subsection"> | 
 | 484 |   <a name="refimpl">Reference Implementation</a> | 
 | 485 | </div> | 
 | 486 | <div class="doc_text"> | 
 | 487 |   <p>The <tt>linux</tt> implementation of the system library will always be the | 
 | 488 |   reference implementation. This means that (a) the concepts defined by the | 
 | 489 |   linux must be identically replicated in the other implementations and (b) the | 
 | 490 |   linux implementation must always be complete (provide implementations for all | 
 | 491 |   concepts).</p> | 
 | 492 | </div> | 
 | 493 |  | 
 | 494 | <!-- *********************************************************************** --> | 
 | 495 |  | 
 | 496 | <hr> | 
 | 497 | <address> | 
 | 498 |   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img | 
 | 499 |   src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a> | 
 | 500 |   <a href="http://validator.w3.org/check/referer"><img | 
 | 501 |   src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a> | 
 | 502 |  | 
 | 503 |   <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br> | 
 | 504 |   <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br> | 
 | 505 |   Last modified: $Date$ | 
 | 506 | </address> | 
 | 507 | </body> | 
 | 508 | </html> |