summaryrefslogtreecommitdiff
path: root/opcodes/configure
diff options
context:
space:
mode:
authorNick Alcock <nick.alcock@oracle.com>2022-09-13 11:26:14 +0100
committerNick Alcock <nick.alcock@oracle.com>2023-01-09 12:50:53 +0000
commit5e72265660fbbfdceafbb1a5a5aad79e835e31cd (patch)
tree6fbdd27aee312c0ad2c4d35b09e631f8129870d9 /opcodes/configure
parentadad40b1df5f7ebffd6c9f90e23a334cf5020db5 (diff)
downloadbinutils-gdb-5e72265660fbbfdceafbb1a5a5aad79e835e31cd.tar.gz
libtool.m4: adjust kludge for ignoring syntax errors
LT_PATH_NM considers an nm to support -p/-B if it returns output that names the path being checked -- but since that is itself nm (usually with a path attached), this can trigger false positives with errors like /usr/bin/nm: unknown option "B" ignored" The test currently avoids this case for the single instance of HP-UX by deleting all but the last line of the output. Keep on doing that (so we don't do more work than we need to), but also explicitly eliminate references to "unknown" and "invalid", with any leading capitalization. This handles Darwin, which says error: nm: invalid argument '-B' Alas, we cannot check the exit code, because some nm's return nonzero exit codes if no symbols are found! Things were also erroneously failing if nm had actual symbols in it, because we were getting output that did not mention nm by name (its usual output). Identify things that look like valid nm output and let it through. PR libctf/29547 * libtool.m4 (LT_PATH_NM): Explicitly filter out 'unknown' and 'invalid' errors when looking for valid output. Let through actual valid nm output too.
Diffstat (limited to 'opcodes/configure')
0 files changed, 0 insertions, 0 deletions