Blame - www/varformats.html - fp2-dev/platform/external/lldb

2011-07-12 22:56:10 +0000

[diff] [blame]

100

<p>This is done by typing

101

102

103

<b>(lldb)</b> type format add -f hex int

104

</td>

105

<table>

106

at the LLDB command line.</p>

Enrico Granata

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>

typedef A B;<br>

typedef B C;<br>

typedef C D;<br>

</code></p>

<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

2011-07-12 22:56:10 +0000

[diff] [blame]

132

133

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

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

2011-07-12 22:56:10 +0000

[diff] [blame]

147

148

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

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

2011-07-12 22:56:10 +0000

[diff] [blame]

185

186

187

<b>(lldb)</b> frame variable counter -f hex

188

</td>

189

<table>

Enrico Granata

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>

<tbody>

<td width="23%"><b>Format name</b></td>

203

<td><b>Abbreviation</b></td>

204

<td><b>Description</b></td>

205

</tr>

206

207

<td><b>default</b></td>

208

209

</td>

210

<td>the default LLDB algorithm is used to pick a

format</td>

</tr>

<td><b>boolean</b></td>

215

216

<td>show this as a true/false boolean, using the

217

customary rule that 0 is false and everything else

is true</td>

</tr>

<td><b>binary</b></td>

222

223

<td>show this as a sequence of bits</td>

224

</tr>

225

226

<td><b>bytes</b></td>

227

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

232

<td><b>bytes with ASCII</b></td>

233

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

240

<td><b>character</b></td>

241

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

247

<td><b>printable character</b></td>

248

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

254

<td><b>complex float</b></td>

255

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

262

<td><b>c-string</b></td>

263

264

<td>show this as a 0-terminated C string</td>

265

</tr>

266

267

<td><b>signed decimal</b></td>

268

269

<td>show this as a signed integer number (this does

270

not perform a cast, it simply shows the bytes as

signed integer)</td>

</tr>

<td><b>enumeration</b></td>

275

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>

</tr>

<td>show this as in hexadecimal notation (this does

285

not perform a cast, it simply shows the bytes as

hex)</td>

</tr>

<td><b>float</b></td>

290

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

296

<td><b>octal</b></td>

297

298

<td>show this in octal notation</td>

299

</tr>

300

301

<td><b>OSType</b></td>

302

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

307

<td><b>unicode16</b></td>

308

309

<td>show this as UTF-16 characters<br>

310

e.g. <code>(float) *c.sp.y = 0xd70a 0x411f</code></td>

311

</tr>

312

313

<td><b>unicode32</b></td>

314

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

320

<td><b>unsigned decimal</b></td>

321

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

327

<td><b>pointer</b></td>

328

329

<td>show this as a native pointer (unless this is

330

really a pointer, the resulting address will

331

probably be invalid)</td>

</tr>

</td>

<td>show this as an array of characters<br>

338

e.g. <code>(char) *c.sp.z = {X}</code></td>

</tr>

int16_t[], uint16_t[]<br>

343

int32_t[], uint32_t[]<br>

344

int64_t[], uint64_t[]<br>

uint128_t[]</b></td>

</td>

<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

355

<td><b>float32[], float64[]</b></td>

356

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

364

<td><b>complex integer</b></td>

365

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

371

<td><b>character array</b></td>

372

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>

</tr>

</tbody>

</table>

</div>

</div>

<h1 class="postheader">type summary</h1>

384

Enrico Granata

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

2011-07-08 02:51:01 +0000

[diff] [blame]

392

<p> <i>before adding a summary...</i><br>

393

394

(i_am_cool) one = {<br>

Enrico Granata

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

2011-07-08 02:51:01 +0000

[diff] [blame]

398

}<br>

399

</code> <br>

400

<i>after adding a summary...</i><br>

401

402

(i_am_cool) one = int = 3, float = 3.14159, char = 69<br>

403

</code> </p>

Enrico Granata

2011-07-19 02:34:21 +0000

[diff] [blame]

404

405

<p>There are two ways to use type summaries: the first one is to bind a <i>

406

summary string</i> to the datatype; the second is to bind a Python script to the

407

datatype. Both options are enabled by the <code>type summary add</code>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

408

command.</p>

Enrico Granata

2011-07-12 22:56:10 +0000

[diff] [blame]

409

<p>In the example, the command we type was:</p>

410

411

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

412

<b>(lldb)</b> type summary add -f "int = ${var.integer}, float = ${var.floating}, char = ${var.character%u}" i_am_cool

Enrico Granata

2011-07-12 22:56:10 +0000

[diff] [blame]

413

</td>

414

<table>

Enrico Granata

2011-07-19 02:34:21 +0000

[diff] [blame]

415

416

<p>Initially, we will focus on summary strings, and then describe the Python binding

417

mechanism.</p>

418

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

</div>

</div>

<h1 class="postheader">Summary Strings</h1>

423

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

424

<p>While you may already have guessed a lot about the format of

425

summary strings from the above example, a detailed description

426

of their format follows.</p>

427

428

<p>Summary strings can contain plain text, control characters and

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

429

special symbols that have access to information about

430

the current object and the overall program state.</p>

431

<p>Normal characters are any text that doesn't contain a <code><b>'{'</b></code>,

432

433

character.</p>

434

<p>Variable names are found in between a <code><b>"${"</b></code>

435

prefix, and end with a <code><b>"}"</b></code> suffix.

436

In other words, a variable looks like <code>"<b>${frame.pc}</b>"</code>.</p>

437

<p>Basically, all the variables described in <a

438

href="formats.html">Frame and Thread Formatting</a>

439

are accepted. Also acceptable are the control characters

440

and scoping features described in that page.

441

Additionally, <code>${var</code> and <code>${*var</code>

442

become acceptable symbols in this scenario.</p>

443

<p>The simplest thing you can do is grab a member variable

444

of a class or structure by typing its <i>expression

445

path</i>. In the previous example, the expression path

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

446

for the floating member is simply <code>.floating</code>.

447

Thus, to ask the summary string to display <code>floating</code>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

448

you would type <code>${var.floating}</code> (<code>${var</code>

449

is a placeholder token replaced with whatever variable

450

is being displayed).</p>

451

<p>If you have code like the following: <br>

452

<code> struct A {<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

453

    int x;<br>

454

    int y;<br>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

455

};<br>

456

struct B {<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

457

    A x;<br>

458

    A y;<br>

459

    int z;<br>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

460

};<br>

461

</code> the expression path for the <code>y</code>

462

member of the <code>x</code> member of an object of

463

type <code>B</code> would be <code>.x.y</code> and you

464

would type <code>${var.x.y}</code> to display it in a

465

summary string for type <code>B</code>. </p>

466

<p>As you could be using a summary string for both

467

displaying objects of type <code>T</code> or <code>T*</code>

468

(unless <code>-p</code> is used to prevent this), the

469

expression paths do not differentiate between <code>.</code>

470

and <code>-></code>, and the above expression path <code>.x.y</code>

471

would be just as good if you were displaying a <code>B*</code>,

472

or even if the actual definition of <code>B</code>

473

were: <code><br>

474

struct B {<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

475

    A *x;<br>

476

    A y;<br>

477

    int z;<br>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

478

};<br>

479

</code> </p>

480

<p>This is unlike the behaviour of <code>frame variable</code>

481

which, on the contrary, will enforce the distinction. As

482

hinted above, the rationale for this choice is that

483

waiving this distinction enables one to write a summary

484

string once for type <code>T</code> and use it for both

485

<code>T</code> and <code>T*</code> instances. As a

486

summary string is mostly about extracting nested

487

members' information, a pointer to an object is just as

488

good as the object itself for the purpose.</p>

489

<p>Of course, you can have multiple entries in one summary

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

490

string, as shown in the previous example.</p>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

491

<p>As you can see, the last expression path also contains

492

a <code>%u</code> symbol which is nowhere to be found

493

in the actual member variable name. The symbol is

494

reminding of a <code>printf()</code> format symbol, and

495

in fact it has a similar effect. If you add a % sign

496

followed by any one format name or abbreviation from the

497

above table after an expression path, the resulting

Enrico Granata

e4e3e2c

2011-07-22 00:16:08 +0000

[diff] [blame^]

498

object will be displyed using the chosen format.</p>

499

500

<p>You can also use some other special format markers, not available

501

for type formatters, but which carry a special meaning when used in this

context:</p>

<tbody>

<td width="23%"><b>Symbol</b></td>

508

<td><b>Description</b></td>

</tr>

<td>Use this object's summary (the default for aggregate types)</td>

</tr>

<td>Use this object's value (the default for non-aggregate types)</td>

</tr>

<td>Use a language-runtime specific description (for C++ this does nothing,

521

for Objective-C it calls the NSPrintForDebugger API)</td>

</tr>

<td>Use this object's location (memory address, register name, ...)</td>

</tr>

</tbody>

</table>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

530

<p>As previously said, pointers and values are treated the

531

same way when getting to their members in an expression

532

path. However, if your expression path leads to a

533

pointer, LLDB will not automatically dereference it. In

534

order to obtain The deferenced value for a pointer, your

535

expression path must start with <code>${*var</code>

536

instead of <code>${var</code>. Because there is no need

537

to dereference pointers along your way, the

538

dereferencing symbol only applies to the result of the

539

whole expression path traversing. <br>

540

e.g. <code> <br>

541

542

(Couple) c = {<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

543

    (SimpleWithPointers) sp = {<br>

544

        (int *) x = 0x00000001001000b0<br>

545

        (float *) y = 0x00000001001000c0<br>

546

        (char *) z = 0x00000001001000d0 "X"<br>

547

    }<br>

548

    (Simple *) s = 0x00000001001000e0<br>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

549

}<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

550

</code><br>

551

552

If one types the following commands:

<b>(lldb)</b> type summary add -f "int = ${*var.sp.x},

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

557

float = ${*var.sp.y}, char = ${*var.sp.z%u}, Simple =

558

${*var.s}" Couple<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

559

<b>(lldb)</b> type summary add -c -p Simple<br>

</td>

the output becomes: <br><code>

564

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

565

566

(Couple) c = int = 9, float = 9.99, char = 88, Simple

567

= (x=9, y=9.99, z='X')<br>

568

</code> </p>

569

<p>Option <code>-c</code> to <code>type summary add</code>

570

tells LLDB not to look for a summary string, but instead

571

to just print a listing of all the object's children on

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

572

one line, as shown in the summary for object Simple.</p>

573

<p> We are using the <code>-p</code> flag here to show that

574

aggregate types can be dereferenced as well as basic types.

575

The following command sequence would work just as well and

576

produce the same output:

577

578

579

<b>(lldb)</b> type summary add -f "int = ${*var.sp.x},

580

float = ${*var.sp.y}, char = ${*var.sp.z%u}, Simple =

581

${var.s}" Couple<br>

582

<b>(lldb)</b> type summary add -c Simple<br>

583

</td>

584

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

</div>

</div>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

588

<h1 class="postheader">Bitfields and array syntax</h1>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

589

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

590

<p>Sometimes, a basic type's value actually represents

591

several different values packed together in a bitfield.

592

With the classical view, there is no way to look at

593

them. Hexadecimal display can help, but if the bits

594

actually span byte boundaries, the help is limited.

595

Binary view would show it all without ambiguity, but is

596

often too detailed and hard to read for real-life

597

scenarios. To cope with the issue, LLDB supports native

598

bitfield formatting in summary strings. If your

599

expression paths leads to a so-called <i>scalar type</i>

600

(the usual int, float, char, double, short, long, long

601

long, double, long double and unsigned variants), you

602

can ask LLDB to only grab some bits out of the value and

603

display them in any format you like. The syntax is

604

similar to that used for arrays, just you can also give

605

a pair of indices separated by a <code>-</code>. <br>

606

e.g. <br>

607

<code> <b>(lldb)</b> fr var float_point<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

608

(float) float_point = -3.14159<br> </code>

609

610

611

<b>(lldb)</b> type summary add -f "Sign: ${var[31]%B}

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

612

Exponent: ${var[30-23]%x} Mantissa: ${var[0-22]%u}"

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

float

</td>

<code>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

618

<b>(lldb)</b> fr var float_point<br>

619

(float) float_point = -3.14159 Sign: true Exponent:

620

0x00000080 Mantissa: 4788184<br>

621

</code> In this example, LLDB shows the internal

622

representation of a <code>float</code> variable by

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

623

extracting bitfields out of a float object.</p>

624

625

<p> As far as the syntax is concerned, it looks

626

much like the normal C array syntax, but also allows you

627

to specify 2 indices, separated by a - symbol (a range).

628

Ranges can be given either with the lower or the higher index

629

first, and range extremes are always included in the bits extracted. </p>

630

631

<p>LLDB also allows to use a similar syntax to display

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

632

array members inside a summary string. For instance, you

633

may want to display all arrays of a given type using a

634

more compact notation than the default, and then just

635

delve into individual array members that prove

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

636

interesting to your debugging task. You can tell

637

LLDB to format arrays in special ways, possibly

638

independent of the way the array members' datatype is formatted. <br>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

639

e.g. <br>

640

<code> <b>(lldb)</b> fr var sarray<br>

641

(Simple [3]) sarray = {<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

    [0] = {<br>

        x = 1<br>

        y = 2<br>

        z = '\x03'<br>

    }<br>

    [1] = {<br>

        x = 4<br>

        y = 5<br>

        z = '\x06'<br>

    }<br>

    [2] = {<br>

        x = 7<br>

        y = 8<br>

        z = '\t'<br>

    }<br>

}<br></code>

<b>(lldb)</b> type summary add -f "${var[].x}" "Simple

[3]"

</td>

<code>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

667

<b>(lldb)</b> fr var sarray<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

668

(Simple [3]) sarray = [1,4,7]<br></code></p>

669

670

<p>The <code>[]</code> symbol amounts to: <i>if <code>var</code>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

671

is an array and I knows its size, apply this summary

672

string to every element of the array</i>. Here, we are

673

asking LLDB to display <code>.x</code> for every

674

element of the array, and in fact this is what happens.

675

If you find some of those integers anomalous, you can

676

then inspect that one item in greater detail, without

677

the array format getting in the way: <br>

678

<code> <b>(lldb)</b> fr var sarray[1]<br>

679

(Simple) sarray[1] = {<br>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

680

    x = 4<br>

681

    y = 5<br>

682

    z = '\x06'<br>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

683

}<br>

684

</code> </p>

685

<p>You can also ask LLDB to only print a subset of the

686

array range by using the same syntax used to extract bit

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

for bitfields:

<b>(lldb)</b> type summary add -f "${var[1-2].x}" "Simple

[3]"

</td>

<code>

<b>(lldb)</b> fr var sarray<br>

696

(Simple [3]) sarray = [4,7]<br></code></p>

697

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

698

<p>The same logic works if you are printing a pointer

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

699

instead of an array, however in this latter case, the empty

700

square brackets operator <code>[]</code>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

701

cannot be used and you need to give exact range limits.</p>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

702

703

<p>In general, LLDB needs the square brackets operator <code>[]</code> in

704

order to handle arrays and pointers correctly, and for pointers it also

705

needs a range. However, a few special cases are defined to make your life easier:

706

<ul>

707

<li>you can print a 0-terminated string (<i>C-string</i>) using the %s format,

708

omitting square brackets, as in:

709

710

711

<b>(lldb)</b> type summary add -f "${var%s}" "char *"

</td>

<table>

This works for <code>char*</code> and <code>char[]</code> objects, and uses the

716

<code>\0</code> terminator

717

when possible to terminate the string, instead of relying on array length.

</li> </ul>

<ul>

<li>anyone of the array formats (<code>int8_t[]</code>,

723

<code>float32{}</code>, ...), and the <code>y</code>, <code>Y</code>

724

and <code>a</code> formats

725

work to print an array of a non-aggregate

726

type, even if square brackets are omitted.

727

728

729

<b>(lldb)</b> type summary add -f "${var%int32_t[]}" "int [10]"

</td>

<table>

</ul>

This feature, however, is not enabled for pointers because there is no

735

way for LLDB to detect the end of the pointed data.

736

<br>

737

This also does not work for other formats (e.g. <code>boolean</code>), and you must

738

specify the square brackets operator to get the expected output.

</p>

</div>

</div>

Enrico Granata

2011-07-19 02:34:21 +0000

[diff] [blame]

744

<h1 class="postheader">Python scripting</h1>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

745

Enrico Granata

2011-07-19 02:34:21 +0000

[diff] [blame]

746

747

<p>Most of the times, summary strings prove good enough for the job of summarizing

748

the contents of a variable. However, as soon as you need to do more than picking

749

some values and rearranging them for display, summary strings stop being an

750

effective tool. This is because summary strings lack the power to actually perform

751

some computation on the value of variables.</p>

752

<p>To solve this issue, you can bind some Python scripting code as a summary for

753

your datatype, and that script has the ability to both extract children variables

754

as the summary strings do and to perform active computation on the extracted

755

values. As a small example, let's say we have a Rectangle class:</p>

<code>

class Rectangle<br/>

{<br/>

private:<br/>

    int height;<br/>

    int width;<br/>

public:<br/>

    Rectangle() : height(3), width(5) {}<br/>

765

    Rectangle(int H) : height(H), width(H*2-1) {}<br/>

766

    Rectangle(int H, int W) : height(H), width(W) {}<br/>

767

768

    int GetHeight() { return height; }<br/>

769

    int GetWidth() { return width; }<br/>

};<br/>

</code>

<p>Summary strings are effective to reduce the screen real estate used by

775

the default viewing mode, but are not effective if we want to display the

776

area, perimeter and length of diagonal of <code>Rectangle</code> objects</p>

777

778

<p>To obtain this, we can simply attach a small Python script to the <code>Rectangle</code>

779

class, as shown in this example:</p>

<b>(lldb)</b> type summary add -P Rectangle<br/>

784

Enter your Python command(s). Type 'DONE' to end.<br/>

785

def function (valobj,dict):<br/>

786

    height_val = valobj.GetChildMemberWithName('height')<br/>

787

    width_val = valobj.GetChildMemberWithName('width')<br/>

788

    height_str = height_val.GetValue()<br/>

789

    width_str = width_val.GetValue()<br/>

790

    height = int(height_str)<br/>

791

    width = int(width_str)<br/>

792

    area = height*width<br/>

793

    perimeter = 2*height + 2*width<br/>

794

    diag = sqrt(height*height+width*width)<br/>

795

    return 'Area: ' + str(area) + ', Perimeter: ' + str(perimeter) + ', Diagonal: ' + str(diag)<br/>

796

    DONE<br/>

797

<b>(lldb)</b> script<br/>

798

Python Interactive Interpreter. To exit, type 'quit()', 'exit()' or Ctrl-D.<br/>

799

>>> from math import sqrt<br/>

800

>>> quit()<br/>

801

<b>(lldb)</b> frame variable<br/>

802

(Rectangle) r1 = Area: 20, Perimeter: 18, Diagonal: 6.40312423743<br/>

803

(Rectangle) r2 = Area: 72, Perimeter: 36, Diagonal: 13.416407865<br/>

804

(Rectangle) r3 = Area: 16, Perimeter: 16, Diagonal: 5.65685424949<br/>

</td>

</table>

<p>In this scenario, you need to enter the interactive interpreter to import the

809

function sqrt() from the math library. As the example shows, everything you enter

810

into the interactive interpreter is saved for you to use it in scripts. This way

811

you can define your own utility functions and use them in your summary scripts if

812

necessary.</p>

813

814

<p>In order to write effective summary scripts, you need to know the LLDB public

815

API, which is the way Python code can access the LLDB object model. For further

816

details on the API you should look at <a href="scripting.html">this page</a>, or at

817

the LLDB <a href="docs.html">doxygen documentation</a> when it becomes available.</p>

818

819

<p>As a brief introduction, your script is encapsulated into a function that is

820

passed two parameters: <code>valobj</code> and <code>dict</code>.</p>

821

822

<p><code>dict</code> is an internal support parameter used by LLDB and you should

823

not use it.<br/><code>valobj</code> is the object encapsulating the actual

824

variable being displayed, and its type is SBValue. The most important thing you can

825

do with an SBValue is retrieve its children objects, by calling

826

<code>GetChildMemberWithName()</code>, passing it the child's name as a string, or ask

827

it for its value, by calling <code>GetValue()</code>, which returns a Python string.

828

</p>

829

830

<p>If you need to delve into several levels of hierarchy, as you can do with summary

831

strings, you must use the method <code>GetValueForExpressionPath()</code>, passing it

832

an expression path just like those you could use for summary strings. However, if you need

833

to access array slices, you cannot do that (yet) via this method call, and you must

834

use <code>GetChildMemberWithName()</code> querying it for the array items one by one.

835

836

<p>Other than interactively typing a Python script there are two other ways for you

837

to input a Python script as a summary:

838

839

<ul>

840

<li> using the -s option to <code>type summary add </code> and typing the script

841

code as an option argument; as in: </ul>

<b>(lldb)</b> type summary add -s "height =

846

int(valobj.GetChildMemberWithName('height').GetValue());width =

847

int(valobj.GetChildMemberWithName('width').GetValue());

848

return 'Area: ' + str(height*width)" Rectangle<br/>

</td>

</table>

<ul>

<li> using the -F option to <code>type summary add </code> and giving the name of a

853

Python function with the correct prototype. Most probably, you will define (or have

854

already defined) the function in the interactive interpreter, or somehow

855

loaded it from a file.

</ul>

</p>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

</div>

</div>

Enrico Granata

2011-07-19 02:34:21 +0000

[diff] [blame]

863

864

<h1 class="postheader">Regular expression typenames</h1>

865

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

866

<p>As you noticed, in order to associate the custom

867

summary string to the array types, one must give the

868

array size as part of the typename. This can long become

869

tiresome when using arrays of different sizes, <code>Simple

870

871

[3]</code>, <code>Simple [9]</code>, <code>Simple

872

[12]</code>, ...</p>

873

<p>If you use the <code>-x</code> option, type names are

874

treated as regular expressions instead of type names.

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

875

This would let you rephrase the above example

876

for arrays of type <code>Simple [3]</code> as: <br>

<b>(lldb)</b> type summary add -f "${var[].x}"

881

-x "Simple \[[0-9]+\]"

</td>

<table>

<code>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

886

<b>(lldb)</b> fr var sarray<br>

887

(Simple [3]) sarray = [1,4,7]<br>

888

</code> The above scenario works for <code>Simple [3]</code>

889

as well as for any other array of <code>Simple</code>

890

objects. </p>

891

<p>While this feature is mostly useful for arrays, you

892

could also use regular expressions to catch other type

893

sets grouped by name. However, as regular expression

894

matching is slower than normal name matching, LLDB will

895

first try to match by name in any way it can, and only

896

when this fails, will it resort to regular expression

897

matching. Thus, if your type has a base class with a

898

cascading summary, this will be preferred over any

899

regular expression match for your type itself.</p>

900

</div>

901

</div>

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

902

903

904

<h1 class="postheader">Named summaries</h1>

905

906

<p>For a given datatype, there may be different meaningful summary

907

representations. However, currently, only one summary can be associated

908

to a given datatype. If you need to temporarily override the association

909

for a variable, without changing the summary string bound to the datatype,

910

you can use named summaries.</p>

911

912

<p>Named summaries work by attaching a name to a summary string when creating

913

it. Then, when there is a need to attach the summary string to a variable, the

914

<code>frame variable</code> command, supports a <code>--summary</code> option

915

that tells LLDB to use the named summary given instead of the default one.</p>

<b>(lldb)</b> type summary add -f "x=${var.integer}" --name NamedSummary

</td>

<table>

(i_am_cool) one = int = 3, float = 3.14159, char = 69<br>

924

<b>(lldb)</b> fr var one --summary NamedSummary<br>

925

(i_am_cool) one = x=3<br>

926

</code> </p>

927

Enrico Granata

f7a9b14

2011-07-15 02:26:42 +0000

[diff] [blame]

928

<p>When defining a named summmary, binding it to one or more types becomes optional.

929

Even if you bind the named summary to a type, and later change the summary string

930

for that type, the named summary will not be changed by that. You can delete

931

named summaries by using the <code>type summary delete</code> command, as if the

932

summary name was the datatype that the summary is applied to</p>

933

934

<p>A summary attached to a variable using the </code>--summary</code> option,

935

has the same semantics that a custom format attached using the <code>-f</code>

936

option has: it stays attached till you attach a new one, or till you let

937

your program run again.</p>

938

Enrico Granata

2011-07-13 00:00:57 +0000

[diff] [blame]

</div>

</div>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

943

944

<h1 class="postheader">Finding summaries 101</h1>

945

946

<p>While the rules for finding an appropriate format for a

947

type are relatively simple (just go through typedef

948

hierarchies), summaries follow a more complicated

949

process in finding the right summary string for a

950

variable. Namely, what happens is:</p>

951

<ul>

952

<li>If there is a summary for the type of the variable,

953

use it</li>

954

<li>If this object is a pointer, and there is a summary

955

for the pointee type that does not skip pointers, use

956

it</li>

957

<li>If this object is a reference, and there is a

958

summary for the pointee type that does not skip

959

references, use it</li>

960

<li>If this object is an Objective-C class with a parent

961

class, look at the parent class (and parent of parent,

962

...)</li>

963

<li>If this object is a C++ class with base classes,

964

look at base classes (and bases of bases, ...)</li>

965

<li>If this object is a C++ class with virtual base

966

classes, look at the virtual base classes (and bases

967

of bases, ...)</li>

968

<li>If this object's type is a typedef, go through

Enrico Granata

2011-07-12 22:56:10 +0000

[diff] [blame]

969

typedef hierarchy (LLDB might not be able to do this if

970

the compiler has not emitted enough information. If the

971

required information to traverse typedef hierarchies is

972

missing, type cascading will not work. The

973

<a href="http://clang.llvm.org/">clang compiler</a>,

974

part of the LLVM project, emits the correct debugging

975

information for LLDB to cascade)</li>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

976

<li>If everything has failed, repeat the above search,

977

looking for regular expressions instead of exact

matches</li>

</ul>

</div>

</div>

<h1 class="postheader">TODOs</h1>

984

985

<ul>

986

<li>There's no way to do multiple dereferencing, and you

987

need to be careful what the dereferencing operation is

988

binding to in complicated scenarios</li>

Enrico Granata

2011-07-08 02:51:01 +0000

[diff] [blame]

989

<li><code>type format add</code> does not support the <code>-x</code>

990

option</li>

Enrico Granata

2011-07-19 02:34:21 +0000

[diff] [blame]

991

<strike><li>Object location cannot be printed in the summary

992

string</li></strike>

Enrico Granata