Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 1 | <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> |
| 2 | <html xmlns="http://www.w3.org/1999/xhtml"> |
| 3 | <head> |
| 4 | <meta http-equiv="Content-Type" content="text/html; |
| 5 | charset=ISO-8859-1"> |
| 6 | <link href="style.css" rel="stylesheet" type="text/css"> |
| 7 | <title>LLDB Homepage</title> |
| 8 | </head> |
| 9 | <body> |
| 10 | <div class="www_title"> The <strong>LLDB</strong> Debugger </div> |
| 11 | <div id="container"> |
| 12 | <div id="content"> |
| 13 | <!--#include virtual="sidebar.incl"--> |
| 14 | <div id="middle"> |
| 15 | <div class="post"> |
| 16 | <h1 class="postheader">Variable display</h1> |
| 17 | <div class="postcontent"> |
| 18 | |
| 19 | <p>LLDB was recently modified to allow users to define custom |
| 20 | formatting options for the variables display.</p> |
| 21 | |
| 22 | <p>Usually, when you type <code>frame variable</code> or |
| 23 | run some <code>expression</code> LLDB will |
| 24 | automatically choose a format to display your results on |
| 25 | a per-type basis, as in the following example:</p> |
| 26 | |
| 27 | <p> <code> <b>(lldb)</b> frame variable -T sp<br> |
| 28 | (SimpleWithPointers) sp = {<br> |
| 29 | (int *) x = 0x0000000100100120<br> |
| 30 | (float *) y = |
| 31 | 0x0000000100100130<br> |
| 32 | (char *) z = |
| 33 | 0x0000000100100140 "6"<br> |
| 34 | }<br> |
| 35 | </code> </p> |
| 36 | |
| 37 | <p>However, in certain cases, you may want to associate a |
| 38 | different format to the display for certain datatypes. |
| 39 | To do so, you need to give hints to the debugger as to |
| 40 | how datatypes should be displayed.<br> |
| 41 | A new <b>type</b> command has been introduced in LLDB |
| 42 | which allows to do just that.<br> |
| 43 | </p> |
| 44 | |
| 45 | <p>Using it you can obtain a format like this one for <code>sp</code>, |
| 46 | instead of the default shown above: </p> |
| 47 | |
| 48 | <p> <code> <b>(lldb)</b> frame variable sp<br> |
| 49 | (SimpleWithPointers) sp = |
| 50 | (x=0x0000000100100120 -> -1, y=0x0000000100100130 |
| 51 | -> -2, z="3")<br> |
| 52 | </code> </p> |
| 53 | |
| 54 | <p>There are two kinds of printing options: <span |
| 55 | style="font-style: italic;">summary</span> and <span |
| 56 | style="font-style: italic;">format</span>. While a |
| 57 | detailed description of both will be given below, one |
| 58 | can briefly say that a summary is mainly used for |
| 59 | aggregate types, while a format is attached to primitive |
| 60 | types.</p> |
| 61 | |
| 62 | <p>To reflect this, the the <b>type</b> command has two |
| 63 | subcommands:<br> |
| 64 | </p> |
| 65 | |
| 66 | <p><code>type format</code></p> |
| 67 | <p><code>type summary</code></p> |
| 68 | |
| 69 | <p>These commands are meant to bind printing options to |
| 70 | types. When variables are printed, LLDB will first check |
| 71 | if custom printing options have been associated to a |
| 72 | variable's type and, if so, use them instead of picking |
| 73 | the default choices.<br> |
| 74 | </p> |
| 75 | |
| 76 | <p>The two commands <code>type format</code> and <code>type |
| 77 | summary</code> each have four subcommands:<br> |
| 78 | </p> |
| 79 | <p><code>add</code>: associates a new printing option to one |
| 80 | or more types</p> |
| 81 | <p><code>delete</code>: deletes an existing association</p> |
| 82 | <p><code>list</code>: provides a listing of all |
| 83 | associations</p> |
| 84 | <p><code>clear</code>: deletes all associations</p> |
| 85 | </div> |
| 86 | </div> |
| 87 | |
| 88 | <div class="post"> |
| 89 | <h1 class="postheader">type format</h1> |
| 90 | <div class="postcontent"> |
| 91 | |
| 92 | <p>Type formats enable you to quickly override the default |
| 93 | format for displaying primitive types (the usual basic |
| 94 | C/C++/ObjC types: int, float, char, ...).</p> |
| 95 | |
| 96 | <p>If for some reason you want all <code>int</code> |
| 97 | variables in your program to print out as hex, you can add |
| 98 | a format to the <code>int</code> type.<br></p> |
| 99 | |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 100 | <p>This is done by typing |
| 101 | <table class="stats" width="620" cellspacing="0"> |
| 102 | <td class="content"> |
| 103 | <b>(lldb)</b> type format add -f hex int |
| 104 | </td> |
| 105 | <table> |
| 106 | at the LLDB command line.</p> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 107 | |
| 108 | <p>The <code>-f</code> option accepts a <a |
| 109 | href="#formatstable">format name</a>, and a list of |
| 110 | types to which you want the new format applied.</p> |
| 111 | |
| 112 | <p>A frequent scenario is that your program has a <code>typedef</code> |
| 113 | for a numeric type that you know represents something |
| 114 | that must be printed in a certain way. Again, you can |
| 115 | add a format just to that typedef by using <code>type |
| 116 | format add</code> with the name alias.</p> |
| 117 | |
| 118 | <p>But things can quickly get hierarchical. Let's say you |
| 119 | have a situation like the following:</p> |
| 120 | |
| 121 | <p><code>typedef int A;<br> |
| 122 | typedef A B;<br> |
| 123 | typedef B C;<br> |
| 124 | typedef C D;<br> |
| 125 | </code></p> |
| 126 | |
| 127 | <p>and you want to show all <code>A</code>'s as hex, all |
| 128 | <code>C'</code>s as pointers and leave the defaults |
| 129 | untouched for other types.</p> |
| 130 | |
| 131 | <p>If you simply type <br> |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 132 | <table class="stats" width="620" cellspacing="0"> |
| 133 | <td class="content"> |
| 134 | <b>(lldb)</b> type format add -f hex A<br> |
| 135 | <b>(lldb)</b> type format add -f pointer C |
| 136 | </td> |
| 137 | <table> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 138 | <br> |
| 139 | values of type <code>B</code> will be shown as hex |
| 140 | and values of type <code>D</code> as pointers.</p> |
| 141 | |
| 142 | <p>This is because by default LLDB <i>cascades</i> |
| 143 | formats through typedef chains. In order to avoid that |
| 144 | you can use the option <code>-C no</code> to prevent |
| 145 | cascading, thus making the two commands required to |
| 146 | achieve your goal:<br> |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 147 | <table class="stats" width="620" cellspacing="0"> |
| 148 | <td class="content"> |
| 149 | <b>(lldb)</b> type format add -C no -f hex A<br> |
| 150 | <b>(lldb)</b> type format add -C no -f pointer C |
| 151 | </td> |
| 152 | <table> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 153 | |
| 154 | <p>Two additional options that you will want to look at |
| 155 | are <code>-p</code> and <code>-r</code>. These two |
| 156 | options prevent LLDB from applying a format for type <code>T</code> |
| 157 | to values of type <code>T*</code> and <code>T&</code> |
| 158 | respectively.</p> |
| 159 | |
| 160 | <p> <code> <b>(lldb)</b> type format add -f float32[] |
| 161 | int<br> |
| 162 | <b>(lldb)</b> fr var pointer *pointer -T<br> |
| 163 | (int *) pointer = {1.46991e-39 1.4013e-45}<br> |
| 164 | (int) *pointer = {1.53302e-42}<br> |
| 165 | <b>(lldb)</b> type format add -f float32[] int -p<br> |
| 166 | <b>(lldb)</b> fr var pointer *pointer -T<br> |
| 167 | (int *) pointer = 0x0000000100100180<br> |
| 168 | (int) *pointer = {1.53302e-42}<br> |
| 169 | </code> </p> |
| 170 | |
| 171 | <p>As the previous example highlights, you will most |
| 172 | probably want to use <code>-p</code> for your formats.</p> |
| 173 | |
| 174 | <p>If you need to delete a custom format simply type <code>type |
| 175 | format delete</code> followed by the name of the type |
| 176 | to which the format applies. To delete ALL formats, use |
| 177 | <code>type format clear</code>. To see all the formats |
| 178 | defined, type <code>type format list</code>.<br> |
| 179 | </p> |
| 180 | |
| 181 | <p>If all you need to do, however, is display one variable |
| 182 | in a custom format, while leaving the others of the same |
| 183 | type untouched, you can simply type:<br> |
| 184 | <br> |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 185 | <table class="stats" width="620" cellspacing="0"> |
| 186 | <td class="content"> |
| 187 | <b>(lldb)</b> frame variable counter -f hex |
| 188 | </td> |
| 189 | <table> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 190 | |
| 191 | <p>This has the effect of displaying the value of <code>counter</code> |
| 192 | as an hexadecimal number, and will keep showing it this |
| 193 | way until you either pick a different format or till you |
| 194 | let your program run again.</p> |
| 195 | |
| 196 | <p>Finally, this is a list of formatting options available |
| 197 | out of |
| 198 | which you can pick:</p><a name="formatstable"></a> |
| 199 | <table border="1"> |
| 200 | <tbody> |
| 201 | <tr valign="top"> |
| 202 | <td width="23%"><b>Format name</b></td> |
| 203 | <td><b>Abbreviation</b></td> |
| 204 | <td><b>Description</b></td> |
| 205 | </tr> |
| 206 | <tr valign="top"> |
| 207 | <td><b>default</b></td> |
| 208 | <td><br> |
| 209 | </td> |
| 210 | <td>the default LLDB algorithm is used to pick a |
| 211 | format</td> |
| 212 | </tr> |
| 213 | <tr valign="top"> |
| 214 | <td><b>boolean</b></td> |
| 215 | <td>B</td> |
| 216 | <td>show this as a true/false boolean, using the |
| 217 | customary rule that 0 is false and everything else |
| 218 | is true</td> |
| 219 | </tr> |
| 220 | <tr valign="top"> |
| 221 | <td><b>binary</b></td> |
| 222 | <td>b</td> |
| 223 | <td>show this as a sequence of bits</td> |
| 224 | </tr> |
| 225 | <tr valign="top"> |
| 226 | <td><b>bytes</b></td> |
| 227 | <td>y</td> |
| 228 | <td>show the bytes one after the other<br> |
| 229 | e.g. <code>(int) s.x = 07 00 00 00</code></td> |
| 230 | </tr> |
| 231 | <tr valign="top"> |
| 232 | <td><b>bytes with ASCII</b></td> |
| 233 | <td>Y</td> |
| 234 | <td>show the bytes, but try to print them as ASCII |
| 235 | characters<br> |
| 236 | e.g. <code>(int *) c.sp.x = 50 f8 bf 5f ff 7f 00 |
| 237 | 00 P.._....</code></td> |
| 238 | </tr> |
| 239 | <tr valign="top"> |
| 240 | <td><b>character</b></td> |
| 241 | <td>c</td> |
| 242 | <td>show the bytes printed as ASCII characters<br> |
| 243 | e.g. <code>(int *) c.sp.x = |
| 244 | P\xf8\xbf_\xff\x7f\0\0</code></td> |
| 245 | </tr> |
| 246 | <tr valign="top"> |
| 247 | <td><b>printable character</b></td> |
| 248 | <td>C</td> |
| 249 | <td>show the bytes printed as printable ASCII |
| 250 | characters<br> |
| 251 | e.g. <code>(int *) c.sp.x = P.._....</code></td> |
| 252 | </tr> |
| 253 | <tr valign="top"> |
| 254 | <td><b>complex float</b></td> |
| 255 | <td>F</td> |
| 256 | <td>interpret this value as the real and imaginary |
| 257 | part of a complex floating-point number<br> |
| 258 | e.g. <code>(int *) c.sp.x = 2.76658e+19 + |
| 259 | 4.59163e-41i</code></td> |
| 260 | </tr> |
| 261 | <tr valign="top"> |
| 262 | <td><b>c-string</b></td> |
| 263 | <td>s</td> |
| 264 | <td>show this as a 0-terminated C string</td> |
| 265 | </tr> |
| 266 | <tr valign="top"> |
| 267 | <td><b>signed decimal</b></td> |
| 268 | <td>i</td> |
| 269 | <td>show this as a signed integer number (this does |
| 270 | not perform a cast, it simply shows the bytes as |
| 271 | signed integer)</td> |
| 272 | </tr> |
| 273 | <tr valign="top"> |
| 274 | <td><b>enumeration</b></td> |
| 275 | <td>E</td> |
| 276 | <td>show this as an enumeration, printing the |
| 277 | value's name if available or the integer value |
| 278 | otherwise<br> |
| 279 | e.g. <code>(enum enumType) val_type = eValue2</code></td> |
| 280 | </tr> |
| 281 | <tr valign="top"> |
| 282 | <td><b>hex</b></td> |
| 283 | <td>x</td> |
| 284 | <td>show this as in hexadecimal notation (this does |
| 285 | not perform a cast, it simply shows the bytes as |
| 286 | hex)</td> |
| 287 | </tr> |
| 288 | <tr valign="top"> |
| 289 | <td><b>float</b></td> |
| 290 | <td>f</td> |
| 291 | <td>show this as a floating-point number (this does |
| 292 | not perform a cast, it simply interprets the bytes |
| 293 | as an IEEE754 floating-point value)</td> |
| 294 | </tr> |
| 295 | <tr valign="top"> |
| 296 | <td><b>octal</b></td> |
| 297 | <td>o</td> |
| 298 | <td>show this in octal notation</td> |
| 299 | </tr> |
| 300 | <tr valign="top"> |
| 301 | <td><b>OSType</b></td> |
| 302 | <td>O</td> |
| 303 | <td>show this as a MacOS OSType<br> |
| 304 | e.g. <code>(float) *c.sp.y = '\n\x1f\xd7\n'</code></td> |
| 305 | </tr> |
| 306 | <tr valign="top"> |
| 307 | <td><b>unicode16</b></td> |
| 308 | <td>U</td> |
| 309 | <td>show this as UTF-16 characters<br> |
| 310 | e.g. <code>(float) *c.sp.y = 0xd70a 0x411f</code></td> |
| 311 | </tr> |
| 312 | <tr valign="top"> |
| 313 | <td><b>unicode32</b></td> |
| 314 | <td><br> |
| 315 | </td> |
| 316 | <td>show this as UTF-32 characters<br> |
| 317 | e.g. <code>(float) *c.sp.y = 0x411fd70a</code></td> |
| 318 | </tr> |
| 319 | <tr valign="top"> |
| 320 | <td><b>unsigned decimal</b></td> |
| 321 | <td>u</td> |
| 322 | <td>show this as an unsigned integer number (this |
| 323 | does not perform a cast, it simply shows the bytes |
| 324 | as unsigned integer)</td> |
| 325 | </tr> |
| 326 | <tr valign="top"> |
| 327 | <td><b>pointer</b></td> |
| 328 | <td>p</td> |
| 329 | <td>show this as a native pointer (unless this is |
| 330 | really a pointer, the resulting address will |
| 331 | probably be invalid)</td> |
| 332 | </tr> |
| 333 | <tr valign="top"> |
| 334 | <td><b>char[]</b></td> |
| 335 | <td><br> |
| 336 | </td> |
| 337 | <td>show this as an array of characters<br> |
| 338 | e.g. <code>(char) *c.sp.z = {X}</code></td> |
| 339 | </tr> |
| 340 | <tr valign="top"> |
| 341 | <td><b>int8_t[], uint8_t[]<br> |
| 342 | int16_t[], uint16_t[]<br> |
| 343 | int32_t[], uint32_t[]<br> |
| 344 | int64_t[], uint64_t[]<br> |
| 345 | uint128_t[]</b></td> |
| 346 | <td><br> |
| 347 | </td> |
| 348 | <td>show this as an array of the corresponding |
| 349 | integer type<br> |
| 350 | e.g.<br> |
| 351 | <code>(int) sarray[0].x = {1 0 0 0}</code><br> |
| 352 | <code>(int) sarray[0].x = {0x00000001}</code></td> |
| 353 | </tr> |
| 354 | <tr valign="top"> |
| 355 | <td><b>float32[], float64[]</b></td> |
| 356 | <td><br> |
| 357 | </td> |
| 358 | <td>show this as an array of the corresponding |
| 359 | floating-point type<br> |
| 360 | e.g. <code>(int *) pointer = {1.46991e-39 |
| 361 | 1.4013e-45}</code></td> |
| 362 | </tr> |
| 363 | <tr valign="top"> |
| 364 | <td><b>complex integer</b></td> |
| 365 | <td>I</td> |
| 366 | <td>interpret this value as the real and imaginary |
| 367 | part of a complex integer number<br> |
| 368 | e.g. <code>(int *) pointer = 1048960 + 1i</code></td> |
| 369 | </tr> |
| 370 | <tr valign="top"> |
| 371 | <td><b>character array</b></td> |
| 372 | <td>a</td> |
| 373 | <td>show this as a character array<br> |
| 374 | e.g. <code>(int *) pointer = |
| 375 | \x80\x01\x10\0\x01\0\0\0</code></td> |
| 376 | </tr> |
| 377 | </tbody> |
| 378 | </table> |
| 379 | </div> |
| 380 | </div> |
| 381 | |
| 382 | <div class="post"> |
| 383 | <h1 class="postheader">type summary</h1> |
| 384 | <div class="postcontent"> |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 385 | <p>Type formats work by showing a different kind of display for |
| 386 | the value of a variable. However, they only work for basic types. |
| 387 | When you want to display a class or struct in a custom format, you |
| 388 | cannot do that using formats.</p> |
| 389 | <p>A different feature, type summaries, works by extracting |
| 390 | information from classes, structures, ... (<i>aggregate types</i>) |
| 391 | and arranging it in a user-defined format, as in the following example:</p> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 392 | <p> <i>before adding a summary...</i><br> |
| 393 | <code> <b>(lldb)</b> fr var -T one<br> |
| 394 | (i_am_cool) one = {<br> |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 395 | (int) integer = 3<br> |
| 396 | (float) floating = 3.14159<br> |
| 397 | (char) character = 'E'<br> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 398 | }<br> |
| 399 | </code> <br> |
| 400 | <i>after adding a summary...</i><br> |
| 401 | <code> <b>(lldb)</b> fr var one<br> |
| 402 | (i_am_cool) one = int = 3, float = 3.14159, char = 69<br> |
| 403 | </code> </p> |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 404 | <p>The way to obtain this effect is to bind a <i>summary string</i> to |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 405 | the datatype using the <code>type summary add</code> |
| 406 | command.</p> |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 407 | <p>In the example, the command we type was:</p> |
| 408 | <table class="stats" width="620" cellspacing="0"> |
| 409 | <td class="content"> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 410 | <b>(lldb)</b> type summary add -f "int = ${var.integer}, float = ${var.floating}, char = ${var.character%u}" i_am_cool |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 411 | </td> |
| 412 | <table> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 413 | </div> |
| 414 | </div> |
| 415 | <div class="post"> |
| 416 | <h1 class="postheader">Summary Strings</h1> |
| 417 | <div class="postcontent"> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 418 | <p>While you may already have guessed a lot about the format of |
| 419 | summary strings from the above example, a detailed description |
| 420 | of their format follows.</p> |
| 421 | |
| 422 | <p>Summary strings can contain plain text, control characters and |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 423 | special symbols that have access to information about |
| 424 | the current object and the overall program state.</p> |
| 425 | <p>Normal characters are any text that doesn't contain a <code><b>'{'</b></code>, |
| 426 | <code><b>'}'</b></code>, <code><b>'$'</b></code>, or <code><b>'\'</b></code> |
| 427 | character.</p> |
| 428 | <p>Variable names are found in between a <code><b>"${"</b></code> |
| 429 | prefix, and end with a <code><b>"}"</b></code> suffix. |
| 430 | In other words, a variable looks like <code>"<b>${frame.pc}</b>"</code>.</p> |
| 431 | <p>Basically, all the variables described in <a |
| 432 | href="formats.html">Frame and Thread Formatting</a> |
| 433 | are accepted. Also acceptable are the control characters |
| 434 | and scoping features described in that page. |
| 435 | Additionally, <code>${var</code> and <code>${*var</code> |
| 436 | become acceptable symbols in this scenario.</p> |
| 437 | <p>The simplest thing you can do is grab a member variable |
| 438 | of a class or structure by typing its <i>expression |
| 439 | path</i>. In the previous example, the expression path |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 440 | for the floating member is simply <code>.floating</code>. |
| 441 | Thus, to ask the summary string to display <code>floating</code> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 442 | you would type <code>${var.floating}</code> (<code>${var</code> |
| 443 | is a placeholder token replaced with whatever variable |
| 444 | is being displayed).</p> |
| 445 | <p>If you have code like the following: <br> |
| 446 | <code> struct A {<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 447 | int x;<br> |
| 448 | int y;<br> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 449 | };<br> |
| 450 | struct B {<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 451 | A x;<br> |
| 452 | A y;<br> |
| 453 | int z;<br> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 454 | };<br> |
| 455 | </code> the expression path for the <code>y</code> |
| 456 | member of the <code>x</code> member of an object of |
| 457 | type <code>B</code> would be <code>.x.y</code> and you |
| 458 | would type <code>${var.x.y}</code> to display it in a |
| 459 | summary string for type <code>B</code>. </p> |
| 460 | <p>As you could be using a summary string for both |
| 461 | displaying objects of type <code>T</code> or <code>T*</code> |
| 462 | (unless <code>-p</code> is used to prevent this), the |
| 463 | expression paths do not differentiate between <code>.</code> |
| 464 | and <code>-></code>, and the above expression path <code>.x.y</code> |
| 465 | would be just as good if you were displaying a <code>B*</code>, |
| 466 | or even if the actual definition of <code>B</code> |
| 467 | were: <code><br> |
| 468 | struct B {<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 469 | A *x;<br> |
| 470 | A y;<br> |
| 471 | int z;<br> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 472 | };<br> |
| 473 | </code> </p> |
| 474 | <p>This is unlike the behaviour of <code>frame variable</code> |
| 475 | which, on the contrary, will enforce the distinction. As |
| 476 | hinted above, the rationale for this choice is that |
| 477 | waiving this distinction enables one to write a summary |
| 478 | string once for type <code>T</code> and use it for both |
| 479 | <code>T</code> and <code>T*</code> instances. As a |
| 480 | summary string is mostly about extracting nested |
| 481 | members' information, a pointer to an object is just as |
| 482 | good as the object itself for the purpose.</p> |
| 483 | <p>Of course, you can have multiple entries in one summary |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 484 | string, as shown in the previous example.</p> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 485 | <p>As you can see, the last expression path also contains |
| 486 | a <code>%u</code> symbol which is nowhere to be found |
| 487 | in the actual member variable name. The symbol is |
| 488 | reminding of a <code>printf()</code> format symbol, and |
| 489 | in fact it has a similar effect. If you add a % sign |
| 490 | followed by any one format name or abbreviation from the |
| 491 | above table after an expression path, the resulting |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 492 | object will be displyed using the chosen format (this is |
| 493 | applicable to non-aggregate types only, with a few |
| 494 | special exceptions discussed below). </p> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 495 | <p>There are two more special format symbols that you can |
| 496 | use only as part of a summary string: <code>%V</code> |
| 497 | and <code>%@</code>. The first one tells LLDB to ignore |
| 498 | summary strings for the type of the object referred by |
| 499 | the expression path and instead print the object's |
| 500 | value. The second is only applicable to Objective-C |
| 501 | classes, and tells LLDB to get the object's description |
| 502 | from the Objective-C runtime. By default, if no format |
| 503 | is provided, LLDB will try to get the object's summary, |
| 504 | and if empty the object's value. If neither can be |
| 505 | obtained, nothing will be displayed.</p> |
| 506 | <p>As previously said, pointers and values are treated the |
| 507 | same way when getting to their members in an expression |
| 508 | path. However, if your expression path leads to a |
| 509 | pointer, LLDB will not automatically dereference it. In |
| 510 | order to obtain The deferenced value for a pointer, your |
| 511 | expression path must start with <code>${*var</code> |
| 512 | instead of <code>${var</code>. Because there is no need |
| 513 | to dereference pointers along your way, the |
| 514 | dereferencing symbol only applies to the result of the |
| 515 | whole expression path traversing. <br> |
| 516 | e.g. <code> <br> |
| 517 | <b>(lldb)</b> fr var -T c<br> |
| 518 | (Couple) c = {<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 519 | (SimpleWithPointers) sp = {<br> |
| 520 | (int *) x = 0x00000001001000b0<br> |
| 521 | (float *) y = 0x00000001001000c0<br> |
| 522 | (char *) z = 0x00000001001000d0 "X"<br> |
| 523 | }<br> |
| 524 | (Simple *) s = 0x00000001001000e0<br> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 525 | }<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 526 | </code><br> |
| 527 | |
| 528 | If one types the following commands: |
| 529 | |
| 530 | <table class="stats" width="620" cellspacing="0"> |
| 531 | <td class="content"> |
| 532 | <b>(lldb)</b> type summary add -f "int = ${*var.sp.x}, |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 533 | float = ${*var.sp.y}, char = ${*var.sp.z%u}, Simple = |
| 534 | ${*var.s}" Couple<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 535 | <b>(lldb)</b> type summary add -c -p Simple<br> |
| 536 | </td> |
| 537 | <table><br> |
| 538 | |
| 539 | the output becomes: <br><code> |
| 540 | |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 541 | <b>(lldb)</b> fr var c<br> |
| 542 | (Couple) c = int = 9, float = 9.99, char = 88, Simple |
| 543 | = (x=9, y=9.99, z='X')<br> |
| 544 | </code> </p> |
| 545 | <p>Option <code>-c</code> to <code>type summary add</code> |
| 546 | tells LLDB not to look for a summary string, but instead |
| 547 | to just print a listing of all the object's children on |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 548 | one line, as shown in the summary for object Simple.</p> |
| 549 | <p> We are using the <code>-p</code> flag here to show that |
| 550 | aggregate types can be dereferenced as well as basic types. |
| 551 | The following command sequence would work just as well and |
| 552 | produce the same output: |
| 553 | <table class="stats" width="620" cellspacing="0"> |
| 554 | <td class="content"> |
| 555 | <b>(lldb)</b> type summary add -f "int = ${*var.sp.x}, |
| 556 | float = ${*var.sp.y}, char = ${*var.sp.z%u}, Simple = |
| 557 | ${var.s}" Couple<br> |
| 558 | <b>(lldb)</b> type summary add -c Simple<br> |
| 559 | </td> |
| 560 | <table><br> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 561 | </div> |
| 562 | </div> |
| 563 | <div class="post"> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 564 | <h1 class="postheader">Bitfields and array syntax</h1> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 565 | <div class="postcontent"> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 566 | <p>Sometimes, a basic type's value actually represents |
| 567 | several different values packed together in a bitfield. |
| 568 | With the classical view, there is no way to look at |
| 569 | them. Hexadecimal display can help, but if the bits |
| 570 | actually span byte boundaries, the help is limited. |
| 571 | Binary view would show it all without ambiguity, but is |
| 572 | often too detailed and hard to read for real-life |
| 573 | scenarios. To cope with the issue, LLDB supports native |
| 574 | bitfield formatting in summary strings. If your |
| 575 | expression paths leads to a so-called <i>scalar type</i> |
| 576 | (the usual int, float, char, double, short, long, long |
| 577 | long, double, long double and unsigned variants), you |
| 578 | can ask LLDB to only grab some bits out of the value and |
| 579 | display them in any format you like. The syntax is |
| 580 | similar to that used for arrays, just you can also give |
| 581 | a pair of indices separated by a <code>-</code>. <br> |
| 582 | e.g. <br> |
| 583 | <code> <b>(lldb)</b> fr var float_point<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 584 | (float) float_point = -3.14159<br> </code> |
| 585 | <table class="stats" width="620" cellspacing="0"> |
| 586 | <td class="content"> |
| 587 | <b>(lldb)</b> type summary add -f "Sign: ${var[31]%B} |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 588 | Exponent: ${var[30-23]%x} Mantissa: ${var[0-22]%u}" |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 589 | float |
| 590 | </td> |
| 591 | <table><br> |
| 592 | |
| 593 | <code> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 594 | <b>(lldb)</b> fr var float_point<br> |
| 595 | (float) float_point = -3.14159 Sign: true Exponent: |
| 596 | 0x00000080 Mantissa: 4788184<br> |
| 597 | </code> In this example, LLDB shows the internal |
| 598 | representation of a <code>float</code> variable by |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 599 | extracting bitfields out of a float object.</p> |
| 600 | |
| 601 | <p> As far as the syntax is concerned, it looks |
| 602 | much like the normal C array syntax, but also allows you |
| 603 | to specify 2 indices, separated by a - symbol (a range). |
| 604 | Ranges can be given either with the lower or the higher index |
| 605 | first, and range extremes are always included in the bits extracted. </p> |
| 606 | |
| 607 | <p>LLDB also allows to use a similar syntax to display |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 608 | array members inside a summary string. For instance, you |
| 609 | may want to display all arrays of a given type using a |
| 610 | more compact notation than the default, and then just |
| 611 | delve into individual array members that prove |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 612 | interesting to your debugging task. You can tell |
| 613 | LLDB to format arrays in special ways, possibly |
| 614 | independent of the way the array members' datatype is formatted. <br> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 615 | e.g. <br> |
| 616 | <code> <b>(lldb)</b> fr var sarray<br> |
| 617 | (Simple [3]) sarray = {<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 618 | [0] = {<br> |
| 619 | x = 1<br> |
| 620 | y = 2<br> |
| 621 | z = '\x03'<br> |
| 622 | }<br> |
| 623 | [1] = {<br> |
| 624 | x = 4<br> |
| 625 | y = 5<br> |
| 626 | z = '\x06'<br> |
| 627 | }<br> |
| 628 | [2] = {<br> |
| 629 | x = 7<br> |
| 630 | y = 8<br> |
| 631 | z = '\t'<br> |
| 632 | }<br> |
| 633 | }<br></code> |
| 634 | |
| 635 | <table class="stats" width="620" cellspacing="0"> |
| 636 | <td class="content"> |
| 637 | <b>(lldb)</b> type summary add -f "${var[].x}" "Simple |
| 638 | [3]" |
| 639 | </td> |
| 640 | <table><br> |
| 641 | |
| 642 | <code> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 643 | <b>(lldb)</b> fr var sarray<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 644 | (Simple [3]) sarray = [1,4,7]<br></code></p> |
| 645 | |
| 646 | <p>The <code>[]</code> symbol amounts to: <i>if <code>var</code> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 647 | is an array and I knows its size, apply this summary |
| 648 | string to every element of the array</i>. Here, we are |
| 649 | asking LLDB to display <code>.x</code> for every |
| 650 | element of the array, and in fact this is what happens. |
| 651 | If you find some of those integers anomalous, you can |
| 652 | then inspect that one item in greater detail, without |
| 653 | the array format getting in the way: <br> |
| 654 | <code> <b>(lldb)</b> fr var sarray[1]<br> |
| 655 | (Simple) sarray[1] = {<br> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 656 | x = 4<br> |
| 657 | y = 5<br> |
| 658 | z = '\x06'<br> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 659 | }<br> |
| 660 | </code> </p> |
| 661 | <p>You can also ask LLDB to only print a subset of the |
| 662 | array range by using the same syntax used to extract bit |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 663 | for bitfields: |
| 664 | <table class="stats" width="620" cellspacing="0"> |
| 665 | <td class="content"> |
| 666 | <b>(lldb)</b> type summary add -f "${var[1-2].x}" "Simple |
| 667 | [3]" |
| 668 | </td> |
| 669 | <table><br> |
| 670 | <code> |
| 671 | <b>(lldb)</b> fr var sarray<br> |
| 672 | (Simple [3]) sarray = [4,7]<br></code></p> |
| 673 | |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 674 | <p>The same logic works if you are printing a pointer |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 675 | instead of an array, however in this latter case, the empty |
| 676 | square brackets operator <code>[]</code> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 677 | cannot be used and you need to give exact range limits.</p> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 678 | |
| 679 | <p>In general, LLDB needs the square brackets operator <code>[]</code> in |
| 680 | order to handle arrays and pointers correctly, and for pointers it also |
| 681 | needs a range. However, a few special cases are defined to make your life easier: |
| 682 | <ul> |
| 683 | <li>you can print a 0-terminated string (<i>C-string</i>) using the %s format, |
| 684 | omitting square brackets, as in: |
| 685 | <table class="stats" width="620" cellspacing="0"> |
| 686 | <td class="content"> |
| 687 | <b>(lldb)</b> type summary add -f "${var%s}" "char *" |
| 688 | </td> |
| 689 | <table> |
| 690 | |
| 691 | This works for <code>char*</code> and <code>char[]</code> objects, and uses the |
| 692 | <code>\0</code> terminator |
| 693 | when possible to terminate the string, instead of relying on array length. |
| 694 | |
| 695 | </li> </ul> |
| 696 | <ul> |
| 697 | |
| 698 | <li>anyone of the array formats (<code>int8_t[]</code>, |
| 699 | <code>float32{}</code>, ...), and the <code>y</code>, <code>Y</code> |
| 700 | and <code>a</code> formats |
| 701 | work to print an array of a non-aggregate |
| 702 | type, even if square brackets are omitted. |
| 703 | <table class="stats" width="620" cellspacing="0"> |
| 704 | <td class="content"> |
| 705 | <b>(lldb)</b> type summary add -f "${var%int32_t[]}" "int [10]" |
| 706 | </td> |
| 707 | <table> |
| 708 | |
| 709 | </ul> |
| 710 | This feature, however, is not enabled for pointers because there is no |
| 711 | way for LLDB to detect the end of the pointed data. |
| 712 | <br> |
| 713 | This also does not work for other formats (e.g. <code>boolean</code>), and you must |
| 714 | specify the square brackets operator to get the expected output. |
| 715 | </p> |
| 716 | </div> |
| 717 | </div> |
| 718 | |
| 719 | <div class="post"> |
| 720 | <h1 class="postheader">Regular expression typenames</h1> |
| 721 | <div class="postcontent"> |
| 722 | </div> |
| 723 | </div> |
| 724 | |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 725 | <p>As you noticed, in order to associate the custom |
| 726 | summary string to the array types, one must give the |
| 727 | array size as part of the typename. This can long become |
| 728 | tiresome when using arrays of different sizes, <code>Simple |
| 729 | |
| 730 | [3]</code>, <code>Simple [9]</code>, <code>Simple |
| 731 | [12]</code>, ...</p> |
| 732 | <p>If you use the <code>-x</code> option, type names are |
| 733 | treated as regular expressions instead of type names. |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 734 | This would let you rephrase the above example |
| 735 | for arrays of type <code>Simple [3]</code> as: <br> |
| 736 | |
| 737 | <table class="stats" width="620" cellspacing="0"> |
| 738 | <td class="content"> |
| 739 | <b>(lldb)</b> type summary add -f "${var[].x}" |
| 740 | -x "Simple \[[0-9]+\]" |
| 741 | </td> |
| 742 | <table> |
| 743 | |
| 744 | <code> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 745 | <b>(lldb)</b> fr var sarray<br> |
| 746 | (Simple [3]) sarray = [1,4,7]<br> |
| 747 | </code> The above scenario works for <code>Simple [3]</code> |
| 748 | as well as for any other array of <code>Simple</code> |
| 749 | objects. </p> |
| 750 | <p>While this feature is mostly useful for arrays, you |
| 751 | could also use regular expressions to catch other type |
| 752 | sets grouped by name. However, as regular expression |
| 753 | matching is slower than normal name matching, LLDB will |
| 754 | first try to match by name in any way it can, and only |
| 755 | when this fails, will it resort to regular expression |
| 756 | matching. Thus, if your type has a base class with a |
| 757 | cascading summary, this will be preferred over any |
| 758 | regular expression match for your type itself.</p> |
| 759 | </div> |
| 760 | </div> |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 761 | |
| 762 | <div class="post"> |
| 763 | <h1 class="postheader">Named summaries</h1> |
| 764 | <div class="postcontent"> |
| 765 | <p>For a given datatype, there may be different meaningful summary |
| 766 | representations. However, currently, only one summary can be associated |
| 767 | to a given datatype. If you need to temporarily override the association |
| 768 | for a variable, without changing the summary string bound to the datatype, |
| 769 | you can use named summaries.</p> |
| 770 | |
| 771 | <p>Named summaries work by attaching a name to a summary string when creating |
| 772 | it. Then, when there is a need to attach the summary string to a variable, the |
| 773 | <code>frame variable</code> command, supports a <code>--summary</code> option |
| 774 | that tells LLDB to use the named summary given instead of the default one.</p> |
| 775 | |
| 776 | <table class="stats" width="620" cellspacing="0"> |
| 777 | <td class="content"> |
| 778 | <b>(lldb)</b> type summary add -f "x=${var.integer}" --name NamedSummary |
| 779 | </td> |
| 780 | <table> |
| 781 | <code> <b>(lldb)</b> fr var one<br> |
| 782 | (i_am_cool) one = int = 3, float = 3.14159, char = 69<br> |
| 783 | <b>(lldb)</b> fr var one --summary NamedSummary<br> |
| 784 | (i_am_cool) one = x=3<br> |
| 785 | </code> </p> |
| 786 | |
Enrico Granata | f7a9b14 | 2011-07-15 02:26:42 +0000 | [diff] [blame^] | 787 | <p>When defining a named summmary, binding it to one or more types becomes optional. |
| 788 | Even if you bind the named summary to a type, and later change the summary string |
| 789 | for that type, the named summary will not be changed by that. You can delete |
| 790 | named summaries by using the <code>type summary delete</code> command, as if the |
| 791 | summary name was the datatype that the summary is applied to</p> |
| 792 | |
| 793 | <p>A summary attached to a variable using the </code>--summary</code> option, |
| 794 | has the same semantics that a custom format attached using the <code>-f</code> |
| 795 | option has: it stays attached till you attach a new one, or till you let |
| 796 | your program run again.</p> |
| 797 | |
Enrico Granata | ede7bdf | 2011-07-13 00:00:57 +0000 | [diff] [blame] | 798 | </div> |
| 799 | </div> |
| 800 | |
| 801 | |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 802 | <div class="post"> |
| 803 | <h1 class="postheader">Finding summaries 101</h1> |
| 804 | <div class="postcontent"> |
| 805 | <p>While the rules for finding an appropriate format for a |
| 806 | type are relatively simple (just go through typedef |
| 807 | hierarchies), summaries follow a more complicated |
| 808 | process in finding the right summary string for a |
| 809 | variable. Namely, what happens is:</p> |
| 810 | <ul> |
| 811 | <li>If there is a summary for the type of the variable, |
| 812 | use it</li> |
| 813 | <li>If this object is a pointer, and there is a summary |
| 814 | for the pointee type that does not skip pointers, use |
| 815 | it</li> |
| 816 | <li>If this object is a reference, and there is a |
| 817 | summary for the pointee type that does not skip |
| 818 | references, use it</li> |
| 819 | <li>If this object is an Objective-C class with a parent |
| 820 | class, look at the parent class (and parent of parent, |
| 821 | ...)</li> |
| 822 | <li>If this object is a C++ class with base classes, |
| 823 | look at base classes (and bases of bases, ...)</li> |
| 824 | <li>If this object is a C++ class with virtual base |
| 825 | classes, look at the virtual base classes (and bases |
| 826 | of bases, ...)</li> |
| 827 | <li>If this object's type is a typedef, go through |
Enrico Granata | 86e7c3e | 2011-07-12 22:56:10 +0000 | [diff] [blame] | 828 | typedef hierarchy (LLDB might not be able to do this if |
| 829 | the compiler has not emitted enough information. If the |
| 830 | required information to traverse typedef hierarchies is |
| 831 | missing, type cascading will not work. The |
| 832 | <a href="http://clang.llvm.org/">clang compiler</a>, |
| 833 | part of the LLVM project, emits the correct debugging |
| 834 | information for LLDB to cascade)</li> |
Enrico Granata | ff78238 | 2011-07-08 02:51:01 +0000 | [diff] [blame] | 835 | <li>If everything has failed, repeat the above search, |
| 836 | looking for regular expressions instead of exact |
| 837 | matches</li> |
| 838 | </ul> |
| 839 | </div> |
| 840 | </div> |
| 841 | <div class="post"> |
| 842 | <h1 class="postheader">TODOs</h1> |
| 843 | <div class="postcontent"> |
| 844 | <ul> |
| 845 | <li>There's no way to do multiple dereferencing, and you |
| 846 | need to be careful what the dereferencing operation is |
| 847 | binding to in complicated scenarios</li> |
| 848 | <li>There is no way to call functions inside summary |
| 849 | strings, not even <code>const</code> ones</li> |
| 850 | <li><code>type format add</code> does not support the <code>-x</code> |
| 851 | option</li> |
| 852 | <li>Object location cannot be printed in the summary |
| 853 | string</li> |
| 854 | </ul> |
| 855 | </div> |
| 856 | </div> |
| 857 | </div> |
| 858 | </div> |
| 859 | </div> |
| 860 | </body> |
| 861 | </html> |