- Don't add path_buf_capacity feature.
- Convert `find_cdb` to early return style, reducing indentation
- Simplify `compute_stamp_hash` for CDB to just hash it's path, if any.
Several Microsoft debuggers (VS, VS Code, WinDbg, CDB, ...) consume the `*.natvis` files we embed into rust `*.pdb` files.
While this only tests CDB, that test coverage should help for all of them.
CHANGES
src\bootstrap
- test.rs: Run CDB debuginfo tests on MSVC targets
src\test\debuginfo
- issue-13213.rs: CDB has trouble with this, skip for now (newly discovered regression?)
- pretty-std.rs: Was ignored, re-enable for CDB only to start with, add CDB tests.
- should-fail.rs: Add CDB tests.
src\tools\compiletest:
- Added "-cdb" option
- Added Mode::DebugInfoCdb ("debuginfo-cdb")
- Added run_debuginfo_cdb_test[_no_opt]
- Renamed Mode::DebugInfoBoth -> DebugInfoGdbLldb ("debuginfo-gdb+lldb") since it's no longer clear what "Both" means.
- Find CDB at the default Win10 SDK install path "C:\Program Files (x86)\Windows Kits\10\Debugger\*\cdb.exe"
- Ignore CDB tests if CDB not found.
ISSUES
- `compute_stamp_hash`: not sure if there's any point in hashing `%ProgramFiles(x86)%`
- `OsString` lacks any `*.natvis` entries (would be nice to add in a followup changelist)
- DSTs (array/string slices) which work in VS & VS Code fail in CDB.
- I've avoided `Mode::DebugInfoAll` as 3 debuggers leads to pow(2,3)=8 possible combinations.
REFERENCE
CDB is not part of the base Visual Studio install, but can be added via the Windows 10 SDK:
https://developer.microsoft.com/en-us/windows/downloads/windows-10-sdk
Installing just "Debugging Tools for Windows" is sufficient.
CDB appears to already be installed on appveyor CI, where this changelist can find it, based on it's use here:
0ffc573110/appveyor.yml (L227)
CDB commands and command line reference:
https://docs.microsoft.com/en-us/windows-hardware/drivers/debugger/debugger-reference
conditionally modify darwin targets to macosx targets with versions
We need this behavior so that Rust LLVM IR objects match the target triple for Clang LLVM IR objects. This matching then convinces the linker that yes, you really can do cross-language LTO with objects from different compilers.
The newly-added tests seem to pass locally on x86_64-unknown-linux-gnu. I haven't done a full test run or tried the new compiler in an cross-language LTO setup yet.
PGO: Add a run-make test that makes sure that PGO profiling data is used by the compiler during optimizations.
From the tests comment section:
```
# This test makes sure that PGO profiling data leads to cold functions being
# marked as `cold` and hot functions with `inlinehint`.
# The test program contains an `if` were actual execution only ever takes the
# `else` branch. Accordingly, we expect the function that is never called to
# be marked as cold.
```
r? @alexcrichton
In JSON output, emit a directive after metadata is generated.
To implement pipelining, Cargo needs to know when metadata generation is
finished. This is done via a new JSON "directive".
Unfortunately, metadata file writing currently occurs very late during
compilation, so pipelining won't produce a speed-up. Moving metadata
file writing earlier will be a follow-up.
r? @alexcrichton
To implement pipelining, Cargo needs to know when metadata generation is
finished. This commit adds code to do that. Unfortunately, metadata file
writing currently occurs very late during compilation, so pipelining
won't produce a speed-up. Moving metadata file writing earlier will be a
follow-up.
The change involves splitting the existing `Emitter::emit` method in
two: `Emitter::emit_diagnostic` and `Emitter::emit_directive`.
The JSON directives look like this:
```
{"directive":"metadata file written: liba.rmeta"}
```
The functionality is behind the `-Z emit-directives` option, and also
requires `--error-format=json`.
Add codegen test for PGO instrumentation.
This PR adds a codegen test that makes sure that LLVM actually generates instrumentation code when we enable PGO instrumentation in `rustc`.
The second commit updates a test case to the new commandline option syntax introduced in #59874. Without the fix the test still works, but it confusingly creates a directory called `test.profraw`, which usually is the name of the _file_ where profiling data is collected.
compiletest normalization: preserve non-JSON lines such as ICEs
Currently, every non-JSON line from stderr gets normalized away when compiletest normalizes the output. In particular, ICEs get normalized to the empty output. That does not seem desirable, so this changes normalization to preserve non-JSON lines instead.
Also see https://github.com/laumann/compiletest-rs/issues/169: because of that bug, Miri currently *looks* green in the toolstate, but some tests ICE. That same bug is likely no longer present in latest compiletest because the error code gets checked separately, but it still seems like a good idea to also make sure that ICEs are considered stderr output:
This change found an accidental user-visible `error!` in CTFE validation (fixed), and a non-deterministic panic when there are two `main` symbols (not fixed, no idea where this comes from). Both got missed before because non-JSON output got ignored.
On musl (and some other platforms), compiletest ends up creating a static rlib
(instead of a dylib) when building 'aux-build' crates.
This commit changes the '--extern-private' path computed by compiletest
to properly take this into account
