ci: Add armhf valgrind to matrix #1857

pull maflcko wants to merge 4 commits into bitcoin-core:master from maflcko:patch-1 changing 2 files +20 −0

maflcko commented at 9:40 AM on May 19, 2026: contributor

This is a hacky rebase of #1433 to add valgrind coverage on armhf for gcc and clang.

However, as explained in #1856, it doesn't work, due to Source and destination overlap in memcpy.
ci: Add armhf valgrind to matrix 856477a8e8
fixup bebc93d8e1
f d750278796
clang 93cb5da20c

in ci/ci.sh:54 in 93cb5da20c

  45 | @@ -45,6 +46,13 @@ if [ -n "${CC+x}" ]; then
  46 |      # The MSVC compiler "cl" doesn't understand "-v"
  47 |      $CC -v || true
  48 |  fi
  49 | +if [ -n "${VALGRIND_EXTRA_PACKAGES+x}" ]; then
  50 | +    # Install an architecture-specific Valgrind build when the job requests it.
  51 | +    # This is used for armhf memcheck jobs on arm64 runners.
  52 | +    # shellcheck disable=SC2086
  53 | +    DEBIAN_FRONTEND=noninteractive apt-get update && \
  54 | +    DEBIAN_FRONTEND=noninteractive apt-get install --no-install-recommends -y $VALGRIND_EXTRA_PACKAGES

maflcko commented at 9:41 AM on May 19, 2026:

If someone picks this up, installation of valgrind:armhf should probably need to move into the dockerfile.

maflcko commented at 10:48 AM on May 19, 2026: contributor

GH will delete the CI log in 90 days, but for reference it was:

tests.log
  ==3547== Memcheck, a memory error detector
  ==3547== Copyright (C) 2002-2024, and GNU GPL'd, by Julian Seward et al.
  ==3547== Using Valgrind-3.24.0 and LibVEX; rerun with -h for copyright info
  ==3547== Command: ./tests
  ==3547== 
  Tests running silently. Use '-log=1' to enable detailed logging
  iterations = 2
  jobs = 0. Sequential execution.
  random seed = 77a8af5c6aa48b86649537af97686fac
  Skipping run_sha256_known_output_tests 1000000 (iteration count too low)
  ==3547== Source and destination overlap in memcpy(0xfeec0288, 0xfeec0288, 148)
  ==3547==    at 0x4881658: memcpy (vg_replace_strmem.c:1150)
  ==3547==    by 0x131C0F: secp256k1_gej_add_var (group_impl.h:541)
  ==3547==    by 0x136B13: test_ge (tests.c:4027)
  ==3547==    by 0x13737D: run_ge (tests.c:4250)
  ==3547==    by 0x109981: run_sequential (unit_test.c:275)
  ==3547==    by 0x109981: tf_run (unit_test.c:464)
  ==3547==    by 0x109981: main (tests.c:8081)
  ==3547==

However, as explained in #1856, it doesn't work, due to Source and destination overlap in memcpy.

Closing for now, let's continue discussion in the issue.

maflcko closed this on May 19, 2026
maflcko deleted the branch on May 19, 2026

Contributors

maflcko

Linked (view graph)

#1433 ci/cirrus: Add ARM32 valgrind tasks #1856 ci: Add ARM32 Valgrind jobs