/
link.rs
1858 lines (1677 loc) · 74.5 KB
/
link.rs
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// Copyright 2012-2014 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
use super::archive::{Archive, ArchiveBuilder, ArchiveConfig, METADATA_FILENAME};
use super::archive;
use super::rpath;
use super::rpath::RPathConfig;
use super::svh::Svh;
use driver::driver::{CrateTranslation, OutputFilenames, Input, FileInput};
use driver::config::NoDebugInfo;
use driver::session::Session;
use driver::config;
use llvm;
use llvm::ModuleRef;
use metadata::common::LinkMeta;
use metadata::{encoder, cstore, filesearch, csearch, loader, creader};
use middle::trans::context::CrateContext;
use middle::trans::common::gensym_name;
use middle::ty;
use util::common::time;
use util::ppaux;
use util::sha2::{Digest, Sha256};
use std::c_str::{ToCStr, CString};
use std::char;
use std::collections::HashSet;
use std::io::{fs, TempDir, Command};
use std::io;
use std::mem;
use std::ptr;
use std::str;
use std::string::String;
use flate;
use serialize::hex::ToHex;
use syntax::abi;
use syntax::ast;
use syntax::ast_map::{PathElem, PathElems, PathName};
use syntax::ast_map;
use syntax::attr::AttrMetaMethods;
use syntax::codemap::Span;
use syntax::parse::token;
// RLIB LLVM-BYTECODE OBJECT LAYOUT
// Version 1
// Bytes Data
// 0..10 "RUST_OBJECT" encoded in ASCII
// 11..14 format version as little-endian u32
// 15..22 size in bytes of deflate compressed LLVM bitcode as
// little-endian u64
// 23.. compressed LLVM bitcode
// This is the "magic number" expected at the beginning of a LLVM bytecode
// object in an rlib.
pub static RLIB_BYTECODE_OBJECT_MAGIC: &'static [u8] = b"RUST_OBJECT";
// The version number this compiler will write to bytecode objects in rlibs
pub static RLIB_BYTECODE_OBJECT_VERSION: u32 = 1;
// The offset in bytes the bytecode object format version number can be found at
pub static RLIB_BYTECODE_OBJECT_VERSION_OFFSET: uint = 11;
// The offset in bytes the size of the compressed bytecode can be found at in
// format version 1
pub static RLIB_BYTECODE_OBJECT_V1_DATASIZE_OFFSET: uint =
RLIB_BYTECODE_OBJECT_VERSION_OFFSET + 4;
// The offset in bytes the compressed LLVM bytecode can be found at in format
// version 1
pub static RLIB_BYTECODE_OBJECT_V1_DATA_OFFSET: uint =
RLIB_BYTECODE_OBJECT_V1_DATASIZE_OFFSET + 8;
#[deriving(Clone, PartialEq, PartialOrd, Ord, Eq)]
pub enum OutputType {
OutputTypeBitcode,
OutputTypeAssembly,
OutputTypeLlvmAssembly,
OutputTypeObject,
OutputTypeExe,
}
pub fn llvm_err(sess: &Session, msg: String) -> ! {
unsafe {
let cstr = llvm::LLVMRustGetLastError();
if cstr == ptr::null() {
sess.fatal(msg.as_slice());
} else {
let err = CString::new(cstr, true);
let err = String::from_utf8_lossy(err.as_bytes());
sess.fatal(format!("{}: {}",
msg.as_slice(),
err.as_slice()).as_slice());
}
}
}
pub fn write_output_file(
sess: &Session,
target: llvm::TargetMachineRef,
pm: llvm::PassManagerRef,
m: ModuleRef,
output: &Path,
file_type: llvm::FileType) {
unsafe {
output.with_c_str(|output| {
let result = llvm::LLVMRustWriteOutputFile(
target, pm, m, output, file_type);
if !result {
llvm_err(sess, "could not write output".to_string());
}
})
}
}
pub mod write {
use super::super::lto;
use super::{write_output_file, OutputType};
use super::{OutputTypeAssembly, OutputTypeBitcode};
use super::{OutputTypeExe, OutputTypeLlvmAssembly};
use super::{OutputTypeObject};
use driver::driver::{CrateTranslation, OutputFilenames};
use driver::config::NoDebugInfo;
use driver::session::Session;
use driver::config;
use llvm;
use llvm::{ModuleRef, TargetMachineRef, PassManagerRef};
use util::common::time;
use syntax::abi;
use std::c_str::ToCStr;
use std::io::{Command};
use libc::{c_uint, c_int};
use std::str;
// On android, we by default compile for armv7 processors. This enables
// things like double word CAS instructions (rather than emulating them)
// which are *far* more efficient. This is obviously undesirable in some
// cases, so if any sort of target feature is specified we don't append v7
// to the feature list.
//
// On iOS only armv7 and newer are supported. So it is useful to
// get all hardware potential via VFP3 (hardware floating point)
// and NEON (SIMD) instructions supported by LLVM.
// Note that without those flags various linking errors might
// arise as some of intrinsics are converted into function calls
// and nobody provides implementations those functions
fn target_feature<'a>(sess: &'a Session) -> &'a str {
match sess.targ_cfg.os {
abi::OsAndroid => {
if "" == sess.opts.cg.target_feature.as_slice() {
"+v7"
} else {
sess.opts.cg.target_feature.as_slice()
}
},
abi::OsiOS if sess.targ_cfg.arch == abi::Arm => {
"+v7,+thumb2,+vfp3,+neon"
},
_ => sess.opts.cg.target_feature.as_slice()
}
}
pub fn run_passes(sess: &Session,
trans: &CrateTranslation,
output_types: &[OutputType],
output: &OutputFilenames) {
let llmod = trans.module;
let llcx = trans.context;
unsafe {
configure_llvm(sess);
if sess.opts.cg.save_temps {
output.with_extension("no-opt.bc").with_c_str(|buf| {
llvm::LLVMWriteBitcodeToFile(llmod, buf);
})
}
let opt_level = match sess.opts.optimize {
config::No => llvm::CodeGenLevelNone,
config::Less => llvm::CodeGenLevelLess,
config::Default => llvm::CodeGenLevelDefault,
config::Aggressive => llvm::CodeGenLevelAggressive,
};
let use_softfp = sess.opts.cg.soft_float;
// FIXME: #11906: Omitting frame pointers breaks retrieving the value of a parameter.
// FIXME: #11954: mac64 unwinding may not work with fp elim
let no_fp_elim = (sess.opts.debuginfo != NoDebugInfo) ||
(sess.targ_cfg.os == abi::OsMacos &&
sess.targ_cfg.arch == abi::X86_64);
// OSX has -dead_strip, which doesn't rely on ffunction_sections
// FIXME(#13846) this should be enabled for windows
let ffunction_sections = sess.targ_cfg.os != abi::OsMacos &&
sess.targ_cfg.os != abi::OsWindows;
let fdata_sections = ffunction_sections;
let reloc_model = match sess.opts.cg.relocation_model.as_slice() {
"pic" => llvm::RelocPIC,
"static" => llvm::RelocStatic,
"default" => llvm::RelocDefault,
"dynamic-no-pic" => llvm::RelocDynamicNoPic,
_ => {
sess.err(format!("{} is not a valid relocation mode",
sess.opts
.cg
.relocation_model).as_slice());
sess.abort_if_errors();
return;
}
};
let code_model = match sess.opts.cg.code_model.as_slice() {
"default" => llvm::CodeModelDefault,
"small" => llvm::CodeModelSmall,
"kernel" => llvm::CodeModelKernel,
"medium" => llvm::CodeModelMedium,
"large" => llvm::CodeModelLarge,
_ => {
sess.err(format!("{} is not a valid code model",
sess.opts
.cg
.code_model).as_slice());
sess.abort_if_errors();
return;
}
};
let tm = sess.targ_cfg
.target_strs
.target_triple
.as_slice()
.with_c_str(|t| {
sess.opts.cg.target_cpu.as_slice().with_c_str(|cpu| {
target_feature(sess).with_c_str(|features| {
llvm::LLVMRustCreateTargetMachine(
t, cpu, features,
code_model,
reloc_model,
opt_level,
true /* EnableSegstk */,
use_softfp,
no_fp_elim,
ffunction_sections,
fdata_sections,
)
})
})
});
// Create the two optimizing pass managers. These mirror what clang
// does, and are by populated by LLVM's default PassManagerBuilder.
// Each manager has a different set of passes, but they also share
// some common passes.
let fpm = llvm::LLVMCreateFunctionPassManagerForModule(llmod);
let mpm = llvm::LLVMCreatePassManager();
// If we're verifying or linting, add them to the function pass
// manager.
let addpass = |pass: &str| {
pass.as_slice().with_c_str(|s| llvm::LLVMRustAddPass(fpm, s))
};
if !sess.no_verify() { assert!(addpass("verify")); }
if !sess.opts.cg.no_prepopulate_passes {
llvm::LLVMRustAddAnalysisPasses(tm, fpm, llmod);
llvm::LLVMRustAddAnalysisPasses(tm, mpm, llmod);
populate_llvm_passes(fpm, mpm, llmod, opt_level,
trans.no_builtins);
}
for pass in sess.opts.cg.passes.iter() {
pass.as_slice().with_c_str(|s| {
if !llvm::LLVMRustAddPass(mpm, s) {
sess.warn(format!("unknown pass {}, ignoring",
*pass).as_slice());
}
})
}
// Finally, run the actual optimization passes
time(sess.time_passes(), "llvm function passes", (), |()|
llvm::LLVMRustRunFunctionPassManager(fpm, llmod));
time(sess.time_passes(), "llvm module passes", (), |()|
llvm::LLVMRunPassManager(mpm, llmod));
// Deallocate managers that we're now done with
llvm::LLVMDisposePassManager(fpm);
llvm::LLVMDisposePassManager(mpm);
// Emit the bytecode if we're either saving our temporaries or
// emitting an rlib. Whenever an rlib is created, the bytecode is
// inserted into the archive in order to allow LTO against it.
if sess.opts.cg.save_temps ||
(sess.crate_types.borrow().contains(&config::CrateTypeRlib) &&
sess.opts.output_types.contains(&OutputTypeExe)) {
output.temp_path(OutputTypeBitcode).with_c_str(|buf| {
llvm::LLVMWriteBitcodeToFile(llmod, buf);
})
}
if sess.lto() {
time(sess.time_passes(), "all lto passes", (), |()|
lto::run(sess, llmod, tm, trans.reachable.as_slice()));
if sess.opts.cg.save_temps {
output.with_extension("lto.bc").with_c_str(|buf| {
llvm::LLVMWriteBitcodeToFile(llmod, buf);
})
}
}
// A codegen-specific pass manager is used to generate object
// files for an LLVM module.
//
// Apparently each of these pass managers is a one-shot kind of
// thing, so we create a new one for each type of output. The
// pass manager passed to the closure should be ensured to not
// escape the closure itself, and the manager should only be
// used once.
fn with_codegen(tm: TargetMachineRef, llmod: ModuleRef,
no_builtins: bool, f: |PassManagerRef|) {
unsafe {
let cpm = llvm::LLVMCreatePassManager();
llvm::LLVMRustAddAnalysisPasses(tm, cpm, llmod);
llvm::LLVMRustAddLibraryInfo(cpm, llmod, no_builtins);
f(cpm);
llvm::LLVMDisposePassManager(cpm);
}
}
let mut object_file = None;
let mut needs_metadata = false;
for output_type in output_types.iter() {
let path = output.path(*output_type);
match *output_type {
OutputTypeBitcode => {
path.with_c_str(|buf| {
llvm::LLVMWriteBitcodeToFile(llmod, buf);
})
}
OutputTypeLlvmAssembly => {
path.with_c_str(|output| {
with_codegen(tm, llmod, trans.no_builtins, |cpm| {
llvm::LLVMRustPrintModule(cpm, llmod, output);
})
})
}
OutputTypeAssembly => {
// If we're not using the LLVM assembler, this function
// could be invoked specially with output_type_assembly,
// so in this case we still want the metadata object
// file.
let ty = OutputTypeAssembly;
let path = if sess.opts.output_types.contains(&ty) {
path
} else {
needs_metadata = true;
output.temp_path(OutputTypeAssembly)
};
with_codegen(tm, llmod, trans.no_builtins, |cpm| {
write_output_file(sess, tm, cpm, llmod, &path,
llvm::AssemblyFile);
});
}
OutputTypeObject => {
object_file = Some(path);
}
OutputTypeExe => {
object_file = Some(output.temp_path(OutputTypeObject));
needs_metadata = true;
}
}
}
time(sess.time_passes(), "codegen passes", (), |()| {
match object_file {
Some(ref path) => {
with_codegen(tm, llmod, trans.no_builtins, |cpm| {
write_output_file(sess, tm, cpm, llmod, path,
llvm::ObjectFile);
});
}
None => {}
}
if needs_metadata {
with_codegen(tm, trans.metadata_module,
trans.no_builtins, |cpm| {
let out = output.temp_path(OutputTypeObject)
.with_extension("metadata.o");
write_output_file(sess, tm, cpm,
trans.metadata_module, &out,
llvm::ObjectFile);
})
}
});
llvm::LLVMRustDisposeTargetMachine(tm);
llvm::LLVMDisposeModule(trans.metadata_module);
llvm::LLVMContextDispose(trans.metadata_context);
llvm::LLVMDisposeModule(llmod);
llvm::LLVMContextDispose(llcx);
if sess.time_llvm_passes() { llvm::LLVMRustPrintPassTimings(); }
}
}
pub fn run_assembler(sess: &Session, outputs: &OutputFilenames) {
let pname = super::get_cc_prog(sess);
let mut cmd = Command::new(pname.as_slice());
cmd.arg("-c").arg("-o").arg(outputs.path(OutputTypeObject))
.arg(outputs.temp_path(OutputTypeAssembly));
debug!("{}", &cmd);
match cmd.output() {
Ok(prog) => {
if !prog.status.success() {
sess.err(format!("linking with `{}` failed: {}",
pname,
prog.status).as_slice());
sess.note(format!("{}", &cmd).as_slice());
let mut note = prog.error.clone();
note.push_all(prog.output.as_slice());
sess.note(str::from_utf8(note.as_slice()).unwrap());
sess.abort_if_errors();
}
},
Err(e) => {
sess.err(format!("could not exec the linker `{}`: {}",
pname,
e).as_slice());
sess.abort_if_errors();
}
}
}
unsafe fn configure_llvm(sess: &Session) {
use std::sync::{Once, ONCE_INIT};
static mut INIT: Once = ONCE_INIT;
// Copy what clang does by turning on loop vectorization at O2 and
// slp vectorization at O3
let vectorize_loop = !sess.opts.cg.no_vectorize_loops &&
(sess.opts.optimize == config::Default ||
sess.opts.optimize == config::Aggressive);
let vectorize_slp = !sess.opts.cg.no_vectorize_slp &&
sess.opts.optimize == config::Aggressive;
let mut llvm_c_strs = Vec::new();
let mut llvm_args = Vec::new();
{
let add = |arg: &str| {
let s = arg.to_c_str();
llvm_args.push(s.as_ptr());
llvm_c_strs.push(s);
};
add("rustc"); // fake program name
if vectorize_loop { add("-vectorize-loops"); }
if vectorize_slp { add("-vectorize-slp"); }
if sess.time_llvm_passes() { add("-time-passes"); }
if sess.print_llvm_passes() { add("-debug-pass=Structure"); }
for arg in sess.opts.cg.llvm_args.iter() {
add((*arg).as_slice());
}
}
INIT.doit(|| {
llvm::LLVMInitializePasses();
// Only initialize the platforms supported by Rust here, because
// using --llvm-root will have multiple platforms that rustllvm
// doesn't actually link to and it's pointless to put target info
// into the registry that Rust cannot generate machine code for.
llvm::LLVMInitializeX86TargetInfo();
llvm::LLVMInitializeX86Target();
llvm::LLVMInitializeX86TargetMC();
llvm::LLVMInitializeX86AsmPrinter();
llvm::LLVMInitializeX86AsmParser();
llvm::LLVMInitializeARMTargetInfo();
llvm::LLVMInitializeARMTarget();
llvm::LLVMInitializeARMTargetMC();
llvm::LLVMInitializeARMAsmPrinter();
llvm::LLVMInitializeARMAsmParser();
llvm::LLVMInitializeMipsTargetInfo();
llvm::LLVMInitializeMipsTarget();
llvm::LLVMInitializeMipsTargetMC();
llvm::LLVMInitializeMipsAsmPrinter();
llvm::LLVMInitializeMipsAsmParser();
llvm::LLVMRustSetLLVMOptions(llvm_args.len() as c_int,
llvm_args.as_ptr());
});
}
unsafe fn populate_llvm_passes(fpm: llvm::PassManagerRef,
mpm: llvm::PassManagerRef,
llmod: ModuleRef,
opt: llvm::CodeGenOptLevel,
no_builtins: bool) {
// Create the PassManagerBuilder for LLVM. We configure it with
// reasonable defaults and prepare it to actually populate the pass
// manager.
let builder = llvm::LLVMPassManagerBuilderCreate();
match opt {
llvm::CodeGenLevelNone => {
// Don't add lifetime intrinsics at O0
llvm::LLVMRustAddAlwaysInlinePass(builder, false);
}
llvm::CodeGenLevelLess => {
llvm::LLVMRustAddAlwaysInlinePass(builder, true);
}
// numeric values copied from clang
llvm::CodeGenLevelDefault => {
llvm::LLVMPassManagerBuilderUseInlinerWithThreshold(builder,
225);
}
llvm::CodeGenLevelAggressive => {
llvm::LLVMPassManagerBuilderUseInlinerWithThreshold(builder,
275);
}
}
llvm::LLVMPassManagerBuilderSetOptLevel(builder, opt as c_uint);
llvm::LLVMRustAddBuilderLibraryInfo(builder, llmod, no_builtins);
// Use the builder to populate the function/module pass managers.
llvm::LLVMPassManagerBuilderPopulateFunctionPassManager(builder, fpm);
llvm::LLVMPassManagerBuilderPopulateModulePassManager(builder, mpm);
llvm::LLVMPassManagerBuilderDispose(builder);
match opt {
llvm::CodeGenLevelDefault | llvm::CodeGenLevelAggressive => {
"mergefunc".with_c_str(|s| llvm::LLVMRustAddPass(mpm, s));
}
_ => {}
};
}
}
/*
* Name mangling and its relationship to metadata. This is complex. Read
* carefully.
*
* The semantic model of Rust linkage is, broadly, that "there's no global
* namespace" between crates. Our aim is to preserve the illusion of this
* model despite the fact that it's not *quite* possible to implement on
* modern linkers. We initially didn't use system linkers at all, but have
* been convinced of their utility.
*
* There are a few issues to handle:
*
* - Linkers operate on a flat namespace, so we have to flatten names.
* We do this using the C++ namespace-mangling technique. Foo::bar
* symbols and such.
*
* - Symbols with the same name but different types need to get different
* linkage-names. We do this by hashing a string-encoding of the type into
* a fixed-size (currently 16-byte hex) cryptographic hash function (CHF:
* we use SHA256) to "prevent collisions". This is not airtight but 16 hex
* digits on uniform probability means you're going to need 2**32 same-name
* symbols in the same process before you're even hitting birthday-paradox
* collision probability.
*
* - Symbols in different crates but with same names "within" the crate need
* to get different linkage-names.
*
* - The hash shown in the filename needs to be predictable and stable for
* build tooling integration. It also needs to be using a hash function
* which is easy to use from Python, make, etc.
*
* So here is what we do:
*
* - Consider the package id; every crate has one (specified with crate_id
* attribute). If a package id isn't provided explicitly, we infer a
* versionless one from the output name. The version will end up being 0.0
* in this case. CNAME and CVERS are taken from this package id. For
* example, github.com/mozilla/CNAME#CVERS.
*
* - Define CMH as SHA256(crateid).
*
* - Define CMH8 as the first 8 characters of CMH.
*
* - Compile our crate to lib CNAME-CMH8-CVERS.so
*
* - Define STH(sym) as SHA256(CMH, type_str(sym))
*
* - Suffix a mangled sym with ::STH@CVERS, so that it is unique in the
* name, non-name metadata, and type sense, and versioned in the way
* system linkers understand.
*/
pub fn find_crate_name(sess: Option<&Session>,
attrs: &[ast::Attribute],
input: &Input) -> String {
use syntax::crateid::CrateId;
let validate = |s: String, span: Option<Span>| {
creader::validate_crate_name(sess, s.as_slice(), span);
s
};
// Look in attributes 100% of the time to make sure the attribute is marked
// as used. After doing this, however, we still prioritize a crate name from
// the command line over one found in the #[crate_name] attribute. If we
// find both we ensure that they're the same later on as well.
let attr_crate_name = attrs.iter().find(|at| at.check_name("crate_name"))
.and_then(|at| at.value_str().map(|s| (at, s)));
match sess {
Some(sess) => {
match sess.opts.crate_name {
Some(ref s) => {
match attr_crate_name {
Some((attr, ref name)) if s.as_slice() != name.get() => {
let msg = format!("--crate-name and #[crate_name] \
are required to match, but `{}` \
!= `{}`", s, name);
sess.span_err(attr.span, msg.as_slice());
}
_ => {},
}
return validate(s.clone(), None);
}
None => {}
}
}
None => {}
}
match attr_crate_name {
Some((attr, s)) => return validate(s.get().to_string(), Some(attr.span)),
None => {}
}
let crate_id = attrs.iter().find(|at| at.check_name("crate_id"))
.and_then(|at| at.value_str().map(|s| (at, s)))
.and_then(|(at, s)| {
from_str::<CrateId>(s.get()).map(|id| (at, id))
});
match crate_id {
Some((attr, id)) => {
match sess {
Some(sess) => {
sess.span_warn(attr.span, "the #[crate_id] attribute is \
deprecated for the \
#[crate_name] attribute");
}
None => {}
}
return validate(id.name, Some(attr.span))
}
None => {}
}
match *input {
FileInput(ref path) => {
match path.filestem_str() {
Some(s) => return validate(s.to_string(), None),
None => {}
}
}
_ => {}
}
"rust-out".to_string()
}
pub fn build_link_meta(sess: &Session, krate: &ast::Crate,
name: String) -> LinkMeta {
let r = LinkMeta {
crate_name: name,
crate_hash: Svh::calculate(&sess.opts.cg.metadata, krate),
};
info!("{}", r);
return r;
}
fn truncated_hash_result(symbol_hasher: &mut Sha256) -> String {
let output = symbol_hasher.result_bytes();
// 64 bits should be enough to avoid collisions.
output.slice_to(8).to_hex().to_string()
}
// This calculates STH for a symbol, as defined above
fn symbol_hash(tcx: &ty::ctxt,
symbol_hasher: &mut Sha256,
t: ty::t,
link_meta: &LinkMeta)
-> String {
// NB: do *not* use abbrevs here as we want the symbol names
// to be independent of one another in the crate.
symbol_hasher.reset();
symbol_hasher.input_str(link_meta.crate_name.as_slice());
symbol_hasher.input_str("-");
symbol_hasher.input_str(link_meta.crate_hash.as_str());
for meta in tcx.sess.crate_metadata.borrow().iter() {
symbol_hasher.input_str(meta.as_slice());
}
symbol_hasher.input_str("-");
symbol_hasher.input_str(encoder::encoded_ty(tcx, t).as_slice());
// Prefix with 'h' so that it never blends into adjacent digits
let mut hash = String::from_str("h");
hash.push_str(truncated_hash_result(symbol_hasher).as_slice());
hash
}
fn get_symbol_hash(ccx: &CrateContext, t: ty::t) -> String {
match ccx.type_hashcodes().borrow().find(&t) {
Some(h) => return h.to_string(),
None => {}
}
let mut symbol_hasher = ccx.symbol_hasher().borrow_mut();
let hash = symbol_hash(ccx.tcx(), &mut *symbol_hasher, t, ccx.link_meta());
ccx.type_hashcodes().borrow_mut().insert(t, hash.clone());
hash
}
// Name sanitation. LLVM will happily accept identifiers with weird names, but
// gas doesn't!
// gas accepts the following characters in symbols: a-z, A-Z, 0-9, ., _, $
pub fn sanitize(s: &str) -> String {
let mut result = String::new();
for c in s.chars() {
match c {
// Escape these with $ sequences
'@' => result.push_str("$SP$"),
'~' => result.push_str("$UP$"),
'*' => result.push_str("$RP$"),
'&' => result.push_str("$BP$"),
'<' => result.push_str("$LT$"),
'>' => result.push_str("$GT$"),
'(' => result.push_str("$LP$"),
')' => result.push_str("$RP$"),
',' => result.push_str("$C$"),
// '.' doesn't occur in types and functions, so reuse it
// for ':' and '-'
'-' | ':' => result.push_char('.'),
// These are legal symbols
'a' .. 'z'
| 'A' .. 'Z'
| '0' .. '9'
| '_' | '.' | '$' => result.push_char(c),
_ => {
let mut tstr = String::new();
char::escape_unicode(c, |c| tstr.push_char(c));
result.push_char('$');
result.push_str(tstr.as_slice().slice_from(1));
}
}
}
// Underscore-qualify anything that didn't start as an ident.
if result.len() > 0u &&
result.as_bytes()[0] != '_' as u8 &&
! char::is_XID_start(result.as_bytes()[0] as char) {
return format!("_{}", result.as_slice());
}
return result;
}
pub fn mangle<PI: Iterator<PathElem>>(mut path: PI,
hash: Option<&str>) -> String {
// Follow C++ namespace-mangling style, see
// http://en.wikipedia.org/wiki/Name_mangling for more info.
//
// It turns out that on OSX you can actually have arbitrary symbols in
// function names (at least when given to LLVM), but this is not possible
// when using unix's linker. Perhaps one day when we just use a linker from LLVM
// we won't need to do this name mangling. The problem with name mangling is
// that it seriously limits the available characters. For example we can't
// have things like &T or ~[T] in symbol names when one would theoretically
// want them for things like impls of traits on that type.
//
// To be able to work on all platforms and get *some* reasonable output, we
// use C++ name-mangling.
let mut n = String::from_str("_ZN"); // _Z == Begin name-sequence, N == nested
fn push(n: &mut String, s: &str) {
let sani = sanitize(s);
n.push_str(format!("{}{}", sani.len(), sani).as_slice());
}
// First, connect each component with <len, name> pairs.
for e in path {
push(&mut n, token::get_name(e.name()).get().as_slice())
}
match hash {
Some(s) => push(&mut n, s),
None => {}
}
n.push_char('E'); // End name-sequence.
n
}
pub fn exported_name(path: PathElems, hash: &str) -> String {
mangle(path, Some(hash))
}
pub fn mangle_exported_name(ccx: &CrateContext, path: PathElems,
t: ty::t, id: ast::NodeId) -> String {
let mut hash = get_symbol_hash(ccx, t);
// Paths can be completely identical for different nodes,
// e.g. `fn foo() { { fn a() {} } { fn a() {} } }`, so we
// generate unique characters from the node id. For now
// hopefully 3 characters is enough to avoid collisions.
static EXTRA_CHARS: &'static str =
"abcdefghijklmnopqrstuvwxyz\
ABCDEFGHIJKLMNOPQRSTUVWXYZ\
0123456789";
let id = id as uint;
let extra1 = id % EXTRA_CHARS.len();
let id = id / EXTRA_CHARS.len();
let extra2 = id % EXTRA_CHARS.len();
let id = id / EXTRA_CHARS.len();
let extra3 = id % EXTRA_CHARS.len();
hash.push_char(EXTRA_CHARS.as_bytes()[extra1] as char);
hash.push_char(EXTRA_CHARS.as_bytes()[extra2] as char);
hash.push_char(EXTRA_CHARS.as_bytes()[extra3] as char);
exported_name(path, hash.as_slice())
}
pub fn mangle_internal_name_by_type_and_seq(ccx: &CrateContext,
t: ty::t,
name: &str) -> String {
let s = ppaux::ty_to_string(ccx.tcx(), t);
let path = [PathName(token::intern(s.as_slice())),
gensym_name(name)];
let hash = get_symbol_hash(ccx, t);
mangle(ast_map::Values(path.iter()), Some(hash.as_slice()))
}
pub fn mangle_internal_name_by_path_and_seq(path: PathElems, flav: &str) -> String {
mangle(path.chain(Some(gensym_name(flav)).move_iter()), None)
}
pub fn get_cc_prog(sess: &Session) -> String {
match sess.opts.cg.linker {
Some(ref linker) => return linker.to_string(),
None => {}
}
// In the future, FreeBSD will use clang as default compiler.
// It would be flexible to use cc (system's default C compiler)
// instead of hard-coded gcc.
// For Windows, there is no cc command, so we add a condition to make it use gcc.
match sess.targ_cfg.os {
abi::OsWindows => "gcc",
_ => "cc",
}.to_string()
}
pub fn get_ar_prog(sess: &Session) -> String {
match sess.opts.cg.ar {
Some(ref ar) => (*ar).clone(),
None => "ar".to_string()
}
}
fn remove(sess: &Session, path: &Path) {
match fs::unlink(path) {
Ok(..) => {}
Err(e) => {
sess.err(format!("failed to remove {}: {}",
path.display(),
e).as_slice());
}
}
}
/// Perform the linkage portion of the compilation phase. This will generate all
/// of the requested outputs for this compilation session.
pub fn link_binary(sess: &Session,
trans: &CrateTranslation,
outputs: &OutputFilenames,
crate_name: &str) -> Vec<Path> {
let mut out_filenames = Vec::new();
for &crate_type in sess.crate_types.borrow().iter() {
if invalid_output_for_target(sess, crate_type) {
sess.bug(format!("invalid output type `{}` for target os `{}`",
crate_type, sess.targ_cfg.os).as_slice());
}
let out_file = link_binary_output(sess, trans, crate_type, outputs,
crate_name);
out_filenames.push(out_file);
}
// Remove the temporary object file and metadata if we aren't saving temps
if !sess.opts.cg.save_temps {
let obj_filename = outputs.temp_path(OutputTypeObject);
if !sess.opts.output_types.contains(&OutputTypeObject) {
remove(sess, &obj_filename);
}
remove(sess, &obj_filename.with_extension("metadata.o"));
}
out_filenames
}
/// Returns default crate type for target
///
/// Default crate type is used when crate type isn't provided neither
/// through cmd line arguments nor through crate attributes
///
/// It is CrateTypeExecutable for all platforms but iOS as there is no
/// way to run iOS binaries anyway without jailbreaking and
/// interaction with Rust code through static library is the only
/// option for now
pub fn default_output_for_target(sess: &Session) -> config::CrateType {
match sess.targ_cfg.os {
abi::OsiOS => config::CrateTypeStaticlib,
_ => config::CrateTypeExecutable
}
}
/// Checks if target supports crate_type as output
pub fn invalid_output_for_target(sess: &Session,
crate_type: config::CrateType) -> bool {
match (sess.targ_cfg.os, crate_type) {
(abi::OsiOS, config::CrateTypeDylib) => true,
_ => false
}
}
fn is_writeable(p: &Path) -> bool {
match p.stat() {
Err(..) => true,
Ok(m) => m.perm & io::UserWrite == io::UserWrite
}
}
pub fn filename_for_input(sess: &Session,
crate_type: config::CrateType,
name: &str,
out_filename: &Path) -> Path {
let libname = format!("{}{}", name, sess.opts.cg.extra_filename);
match crate_type {
config::CrateTypeRlib => {
out_filename.with_filename(format!("lib{}.rlib", libname))
}
config::CrateTypeDylib => {
let (prefix, suffix) = match sess.targ_cfg.os {
abi::OsWindows => (loader::WIN32_DLL_PREFIX, loader::WIN32_DLL_SUFFIX),
abi::OsMacos => (loader::MACOS_DLL_PREFIX, loader::MACOS_DLL_SUFFIX),
abi::OsLinux => (loader::LINUX_DLL_PREFIX, loader::LINUX_DLL_SUFFIX),
abi::OsAndroid => (loader::ANDROID_DLL_PREFIX, loader::ANDROID_DLL_SUFFIX),
abi::OsFreebsd => (loader::FREEBSD_DLL_PREFIX, loader::FREEBSD_DLL_SUFFIX),
abi::OsDragonfly => (loader::DRAGONFLY_DLL_PREFIX, loader::DRAGONFLY_DLL_SUFFIX),
abi::OsiOS => unreachable!(),
};
out_filename.with_filename(format!("{}{}{}",
prefix,
libname,
suffix))
}
config::CrateTypeStaticlib => {
out_filename.with_filename(format!("lib{}.a", libname))
}
config::CrateTypeExecutable => {
match sess.targ_cfg.os {
abi::OsWindows => out_filename.with_extension("exe"),
abi::OsMacos |
abi::OsLinux |
abi::OsAndroid |
abi::OsFreebsd |
abi::OsDragonfly |
abi::OsiOS => out_filename.clone(),
}
}
}
}
fn link_binary_output(sess: &Session,
trans: &CrateTranslation,
crate_type: config::CrateType,
outputs: &OutputFilenames,
crate_name: &str) -> Path {
let obj_filename = outputs.temp_path(OutputTypeObject);