For compiletests, use the linker if present. #109544
Conversation
|
(rustbot has picked a reviewer for you, use r? to override) |
|
rustbot
added
A-testsuite
|
Hey! It looks like you've submitted a new PR for the library teams! If this PR contains changes to any Examples of
|
|
I suspect this fix is too crude, but I'm not familiar enough with the code to do better. Nevertheless, I wanted to document the problem and propose something concrete. |
albertlarsan68
removed
the
T-libs
|
Hello. Do you have any feedback or suggestions on this? |
|
Thanks for the PR and sorry for the delay. |
|
📌 Commit 64225f5a8d90ebef1e3962b1063af3ff2f45dd2c has been approved by It is now in the queue for this repository. |
bors
added
S-waiting-on-bors
|
Maybe squash the commits first so that there is no submodule change ? |
|
My bad. |
bors
added
S-waiting-on-author
|
Thanks for the review. I'm actually not sure this is ready to be committed. The reason is that there is code elsewhere that adds |
When running x.py test tests/ui with a custom linker script, we found
that 204 tests actually invoke the default linker (/usr/bin/ld) instead
of the linker script. Depending on the system linker, this will cause
those tests to fail. Many of the failing tests (196) use
"// aux-build:..." directives. One such example is
tests/ui/annotate-snippet/multispan.rs.
In cases where there were failures, we ran with -v and confirmed that we
saw "linker: Some(<path deleted>)" when the compiletest configuration
was dumped. Nevertheless, the rustc command lines for the failing tests
did not contain -Clinker=XXX, resulting in errors like the following:
= note: /usr/bin/ld: /path/to/libunwind.a(UnwindLevel1-gcc-ext.o): unrecognized relocation (0x2a) in section `.text._Unwind_GetDataRelBase'
/usr/bin/ld: final link failed: Bad value
collect2: error: ld returned 1 exit status
|
@jfgoog: 🔑 Insufficient privileges: not in try users |
|
Could not assign reviewer from: |
|
Just following up again. My original thinking with this CL was that if we went to the trouble of specifying a custom linker, we should use it when running the tests. But I'm not sure if this is excessive, or counter to the intention of allowing a linker to be passed to the test runner. What are your thoughts? |
|
Following up from chat in t-infra/bootstrap: Current theory is that the host could have a custom linker as well, and there is no way to pass that from the bootstrap code, only the target linker. So the logic for |
|
Superseded by #110018 |
When running
x.py test tests/uiwith a custom linker script, we found that 204 tests actually invoke the default linker (/usr/bin/ld) instead of the linker script. Depending on the system linker, this will cause those tests to fail. Many of the failing tests (196) use// aux-build:...directives. One such example is tests/ui/annotate-snippet/multispan.rs.In cases where there were failures, we ran with -v and confirmed that we saw "
linker: Some(<path deleted>)" when the compiletest configuration was dumped. Nevertheless, the rustc command lines for the failing tests did not contain-Clinker=XXX, resulting in errors like the following: