Menu
▾
▴
#496 elfcopy/strip fails a number of GNU binutils tests
It's possible to run the GNU binutils test suite against ELF Tool Chain elfcopy/strip, and it produces a large number of failures. I've submitted tickets for a couple of these relating to SHT_GROUP as that's a current sticking point on FreeBSD, but I am submitting this ticket to track overall progress.
It's simple to run:
cd binutils/testsuite/binutils-all
runtest OBJCOPY=path to elfcopy -a objcopy.exp
It may be that we're not interested in implementing some of these options, but it looks like there are legitimate bugs in supported functionality here.
The output I see:
feynman% runtest OBJCOPY=/tank/emaste/src/elftoolchain/elfcopy/elfcopy -a objcopy.exp
WARNING: Couldn't find the global config file.
WARNING: No tool specified
Test Run By emaste on Wed May 20 13:09:10 2015
Native configuration is amd64-unknown-freebsd10.1
=== tests ===
Schedule of variations:
unix
Running target unix
Using /usr/local/share/dejagnu/baseboards/unix.exp as board description file for target.
Using /usr/local/share/dejagnu/config/unix.exp as generic interface file for target.
Using ./../config/default.exp as tool-and-target-specific interface file.
Running ./objcopy.exp ...
Version /tank/emaste/src/elftoolchain/elfcopy/elfcopy (no version number)
PASS: objcopy (simple copy)
PASS: objcopy -O srec
PASS: objcopy --set-start
PASS: objcopy --adjust-start
FAIL: objcopy --adjust-vma
FAIL: objcopy --adjust-section-vma +
FAIL: objcopy --adjust-section-vma =
PASS: strip preserving OS/ABI
PASS: strip
PASS: strip with saving a symbol
PASS: simple objcopy of executable
PASS: run objcopy of executable
PASS: run stripped executable preserving OS/ABI
PASS: run stripped executable
PASS: run stripped executable with saving a symbol
PASS: keep only debug data
FAIL: simple objcopy of debug data
FAIL: objcopy (ELF unknown section type)
PASS: objcopy (ELF group)
PASS: objcopy (ELF group)
PASS: objcopy (ELF group)
PASS: objcopy (ELF group)
FAIL: copy removing group member
PASS: copy removing all group member
FAIL: copy with setting section flags 1
FAIL: add notes section
FAIL: copy with setting section flags 2
FAIL: copy with setting section flags 3
FAIL: strip --strip-unneeded on common symbol
PASS: strip with section group 1
FAIL: strip with section group 2
UNSUPPORTED: strip empty file
PASS: strip with section group 4
FAIL: strip with section group 5
PASS: strip with section group 6
FAIL: strip with section group 7
PASS: strip with section group 8
FAIL: strip with section group 9
ERROR: .//unique.s: assembly failed
UNRESOLVED: strip on STB_GNU_UNIQUE
PASS: strip -g empty file
PASS: --localize-hidden test 1
FAIL: unordered .debug_info references to .debug_ranges
UNSUPPORTED: unordered .debug_info references to .debug_ranges
PASS: objcopy add-section
FAIL: objcopy add-empty-section
FAIL: objcopy on sections with SHF_EXCLUDE
FAIL: strip --strip-unneeded on sections with SHF_EXCLUDE
PASS: --localize-hidden test 2
=== Summary ===
# of expected passes 26
# of unexpected failures 19
# of unresolved testcases 1
# of unsupported tests 2
Diff:
It's really good to know we can use binutils testsuite on our tools!
I'll try to track down some of these failures later and post updates here.
It looks like the tracker stripped out some of the markup above - it should be something like:
% runtest STRIP=/path/to/elfcopy/strip OBJCOPY=/path/to/elfcopy/elfcopy -a objcopy.expand indeed, at [r3222] I see these section group tests passing:
Thanks!
Related
Commit: [r3222]
For reference, at [r3222]:
Related
Commit: [r3222]
Diff:
Test results from a test run on newer binutils-gdb head (00acd688caf385f38493ebd8d0a7b5f94d501a79):