string_intrinsic (Subprocess aborted) failed

[Pranavchiku]

Description

After building when I run tests using cmake --build build --target test

% cmake --build build --target test
.....
99% tests passed, 1 tests failed out of 286

Label Time Summary:
quadruple_precision    =   0.13 sec*proc (2 tests)

Total Test time (real) =   3.13 sec

The following tests did not run:
	  1 - always_skip (Skipped)
	101 - check4 (Skipped)

The following tests FAILED:
	 45 - string_intrinsic (Subprocess aborted)
Errors while running CTest
Output from these tests are in: /Users/pranavchiku/repos/stdlib-fortran-lang/build/Testing/Temporary/LastTest.log
Use "--rerun-failed --output-on-failure" to re-run the failed cases verbosely.
make: *** [test] Error 8

Expected Behaviour

Tests shall pass.

Version of stdlib

b8fbb3c

Platform and Architecture

MacOS

Additional Information

No response

[jvdp1]

Thank you for reporting this issue. could you rerun the tests with "--rerun failed --output-on-failure" to get more information on which tests are actually failing, please?

[jvdp1]

Could your issue be related with the PR #736 ?

[wpbonelli]

I see the same on macOS 13.6.1, aarch64-apple-darwin22.3.0, GCC 13.2.0. It is indeed the move test that fails, in particular move to self:

stdlib/test/string/test_string_intrinsic.f90

Line 717 in b8fbb3c

call move(from_string, from_string)

[jvdp1]

Thank you @wpbonelli for testing. This issue should have been solved with #736 . Question: is it an issue with the compiler or with the code?
As I don't have a macOS, could you investigate what the problem is, please?

[wpbonelli]

It may be related to apple silicon — @Pranavchiku has not reported arch but presumably CI would fail here if it was universal on mac. I will investigate.

edit: confirmed the test passes on an Intel Mac with gcc 13.2

[wpbonelli]

At this point it seems like a compiler problem. The issue does not reproduce in a minimal test https://github.com/wpbonelli/move-test on arm64. My next step is to add a test-drive harness and see if it still succeeds.

I guess I will try to systematically bring the test project to equivalence with stdlib_string_type.f90 to track it down

[certik]

On Apple M1 Max with GFortran 11.3.0 I get:

$ ./test/string/test_string_intrinsic 
# Testing: string-intrinsic
  Starting lgt ... (1/17)
       ... lgt [PASSED]
  Starting llt ... (2/17)
       ... llt [PASSED]
  Starting lge ... (3/17)
       ... lge [PASSED]
  Starting lle ... (4/17)
       ... lle [PASSED]
  Starting trim ... (5/17)
       ... trim [PASSED]
  Starting len ... (6/17)
       ... len [PASSED]
  Starting len_trim ... (7/17)
       ... len_trim [PASSED]
  Starting adjustl ... (8/17)
       ... adjustl [PASSED]
  Starting adjustr ... (9/17)
       ... adjustr [PASSED]
  Starting scan ... (10/17)
       ... scan [PASSED]
  Starting verify ... (11/17)
       ... verify [PASSED]
  Starting repeat ... (12/17)
       ... repeat [PASSED]
  Starting index ... (13/17)
       ... index [PASSED]
  Starting char ... (14/17)
       ... char [PASSED]
  Starting ichar ... (15/17)
       ... ichar [PASSED]
  Starting iachar ... (16/17)
       ... iachar [PASSED]
  Starting move ... (17/17)
test_string_intrinsic(96879,0x102dd8580) malloc: *** error for object 0x6000006982c0: pointer being freed was not allocated
test_string_intrinsic(96879,0x102dd8580) malloc: *** set a breakpoint in malloc_error_break to debug

Program received signal SIGABRT: Process abort signal.

Backtrace for this error:
#0  0x102fe7c07
#1  0x102fe6cc3
#2  0x18f3244e3
#3  0x18f30ceaf
#4  0x18f24a313
#5  0x18f12fa1b
#6  0x18f13305f
#7  0x18f121fa7
#8  0x102be684f
#9  0x102bf19bb
#10  0x102bf7cc3
#11  0x102be8a4b
#12  0x102bf9e0f
zsh: abort      ./test/string/test_string_intrinsic