https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83760
Bug ID: 83760 Summary: [8 Regression] [SH] ICE in maybe_record_trace_start building glibc tst-copy_file_range.c Product: gcc Version: 8.0 Status: UNCONFIRMED Keywords: ice-on-valid-code Severity: normal Priority: P3 Component: target Assignee: unassigned at gcc dot gnu.org Reporter: jsm28 at gcc dot gnu.org CC: law at gcc dot gnu.org Target Milestone: --- Target: sh4eb-linux-gnu Created attachment 43083 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=43083&action=edit preprocessed source Compile the attached test (from the glibc testsuite) with -g -O2 for sh4eb-linux-gnu, GCC trunk. I see an ICE: during RTL pass: dwarf2 tst-copy_file_range.c: In function 'simple_file_copy': tst-copy_file_range.c:246:1: internal compiler error: in maybe_record_trace_start, at dwarf2cfi.c:2344 0x79ad31 maybe_record_trace_start /scratch/jmyers/glibc/many8/src/gcc/gcc/dwarf2cfi.c:2344 0x79b023 create_trace_edges /scratch/jmyers/glibc/many8/src/gcc/gcc/dwarf2cfi.c:2440 0x79ec6a scan_trace /scratch/jmyers/glibc/many8/src/gcc/gcc/dwarf2cfi.c:2653 0x79f8aa create_cfi_notes /scratch/jmyers/glibc/many8/src/gcc/gcc/dwarf2cfi.c:2679 0x79f8aa execute_dwarf2_frame /scratch/jmyers/glibc/many8/src/gcc/gcc/dwarf2cfi.c:3037 0x79f8aa execute /scratch/jmyers/glibc/many8/src/gcc/gcc/dwarf2cfi.c:3525 Please submit a full bug report, with preprocessed source if appropriate. Please include the complete backtrace with any bug report. See <https://gcc.gnu.org/bugs/> for instructions. This only appeared recently in glibc testing (where the exact sources being compiled would have varied over time), but if I fix the .i file used in testing, my bisection says this appeared with r255387. Which makes no real sense, so I'm supposing that actually exposed a latent bug elsewhere. It's entirely possible the issue is really the same as bug 78459, bug 80863 or both of them (both are SH bugs reporting ICEs in maybe_record_trace_start which later mysteriously disappeared). Note: according to the logs from my glibc bot the problem only appeared for sh4eb-linux-gnu, not for little-endian or soft-float, which might also be an indication of a latent bug being more or less randomly perturbed by other variations.