Re: testsuite regressions on trunk, compared to 2.28

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Wed, Jun 21, 2017 at 3:18 PM, Matthias Klose <doko@ubuntu.com> wrote:
> On 21.06.2017 23:08, H.J. Lu wrote:
>> On Wed, Jun 21, 2017 at 10:48 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> On Wed, Jun 21, 2017 at 8:57 AM, Matthias Klose <doko@ubuntu.com> wrote:
>>>> Preparing for the next binutils release, I did a test rebuild on some
>>>> archictures, and comparing test results with those from binutils 2.28,
>>>> configured with
>>>>
>>>> --enable-shared --enable-plugins --enable-threads --with-system-zlib
>>>> --prefix=/usr --enable-deterministic-archives
>>>> --disable-compressed-debug-sections --enable-new-dtags
>>>>
>>>> binutils 2.28 was configured without --enable-new-dtags
>>>>
>>>> I see the following regressions:
>>>>
>>>> amd64:
>>>> Test results, compared with installed binutils:
>>>> W: [ld-x86-64/x86-64.exp] REGRESSION (PASS -> FAIL): Run pr19031
>>>> W: [ld-x86-64/no-plt.exp] REGRESSION (PASS -> FAIL): Run pr20253-2f
>>>> W: [ld-elf/shared.exp] REGRESSION (PASS -> FAIL): Run dl6a7 with
>>>> --dynamic-list-data -Bsymbolic and dlopen on libdl6a.so
>>>> W: [ld-elf/shared.exp] REGRESSION (PASS -> FAIL): Run dl6a6 with
>>>> --dynamic-list-cpp-new -Bsymbolic-functions and dlopen on libdl6a.so
>>>> W: [ld-elf/shared.exp] REGRESSION (PASS -> FAIL): Run dl6a5 with
>>>> -Bsymbolic-functions --dynamic-list-cpp-new and dlopen on libdl6a.so
>>>> W: [ld-elf/shared.exp] REGRESSION (PASS -> FAIL): Run dl6a2 with
>>>> -Bsymbolic-functions and dlopen on libdl6a.so
>>>> W: [ld-srec/srec.exp] REGRESSION (PASS -> FAIL): S-records
>>>> W: [ld-srec/srec.exp] REGRESSION (PASS -> FAIL): S-records with constructors
>>>> I: [ld-elf/elf.exp] progression (FAIL -> PASS): ld-elf/rpath-1
>>>> I: [ld-elf/elf.exp] progression (FAIL -> PASS): ld-elf/now-3
>>>> I: [ld-elf/elf.exp] progression (FAIL -> PASS): ld-elf/now-4
>>>> I: [ld-elf/elf.exp] progression (FAIL -> PASS): ld-elf/rpath-2
>>>> W: [ld-scripts/crossref.exp] REGRESSION (PASS -> FAIL): NOCROSSREFS 3
>>>> W: [ld-size/size.exp] REGRESSION (PASS -> FAIL): Run size-5a
>>>> W: [ld-size/size.exp] REGRESSION (PASS -> FAIL): Run size-8
>>>> W: [ld-size/size.exp] REGRESSION (PASS -> FAIL): Run size-6
>>>> W: [ld-size/size.exp] REGRESSION (PASS -> FAIL): Run size-5b
>>>> W: [ld-size/size.exp] REGRESSION (PASS -> FAIL): Run size-4a
>>>> W: [ld-size/size.exp] REGRESSION (PASS -> FAIL): Run size-4b
>>>> 15 REGRESSIONS (0.47%).
>>>> 4 progressions (0.12%).
>>>> 3216 tests: 3129 pass (97.29%), 16 fail (0.50%), 57 xfail (1.77%) 1 untested
>>>> (0.03%).
>>>>
>>>> i386:
>>>> Test results, compared with installed binutils:
>>>> I: [ld-elf/elf.exp] progression (FAIL -> PASS): ld-elf/now-4
>>>> I: [ld-elf/elf.exp] progression (FAIL -> PASS): ld-elf/rpath-2
>>>> I: [ld-elf/elf.exp] progression (FAIL -> PASS): ld-elf/rpath-1
>>>> I: [ld-elf/elf.exp] progression (FAIL -> PASS): ld-elf/now-3
>>>> W: [ld-scripts/crossref.exp] REGRESSION (PASS -> FAIL): NOCROSSREFS_TO 3
>>>> W: [ld-scripts/crossref.exp] REGRESSION (PASS -> FAIL): NOCROSSREFS_TO 4
>>>> W: [ld-scripts/crossref.exp] REGRESSION (PASS -> FAIL): NOCROSSREFS 3
>>>> W: [ld-shared/shared.exp] REGRESSION (PASS -> FAIL): shared (non PIC, load offset)
>>>> W: [ld-shared/shared.exp] REGRESSION (PASS -> FAIL): shared (PIC main, non PIC so)
>>>> W: [ld-shared/shared.exp] REGRESSION (PASS -> FAIL): shared (non PIC)
>>>> W: [ld-elfvsb/elfvsb.exp] REGRESSION (PASS -> FAIL): visibility (protected) (PIC
>>>> main, non PIC so)
>>>> W: [ld-elfvsb/elfvsb.exp] REGRESSION (PASS -> FAIL): visibility (hidden_weak)
>>>> (non PIC, load offset)
>>>> W: [ld-elfvsb/elfvsb.exp] REGRESSION (PASS -> FAIL): visibility (protected) (non
>>>> PIC, load offset)
>>>> W: [ld-elfvsb/elfvsb.exp] REGRESSION (PASS -> FAIL): visibility (hidden_weak)
>>>> (non PIC)
>>>> W: [ld-elfvsb/elfvsb.exp] REGRESSION (PASS -> FAIL): visibility (hidden_weak)
>>>> (PIC main)
>>>> W: [ld-elfvsb/elfvsb.exp] REGRESSION (PASS -> FAIL): visibility (hidden_weak)
>>>> (PIC main, non PIC so)
>>>> W: [ld-elfvsb/elfvsb.exp] REGRESSION (PASS -> FAIL): visibility (protected) (non
>>>> PIC)
>>>> W: [ld-elfvsb/elfvsb.exp] REGRESSION (PASS -> FAIL): visibility (hidden_weak)
>>>> W: [ld-elf/shared.exp] REGRESSION (PASS -> FAIL): Run dl6a6 with
>>>> --dynamic-list-cpp-new -Bsymbolic-functions and dlopen on libdl6a.so
>>>> W: [ld-elf/shared.exp] REGRESSION (PASS -> FAIL): Run dl6a7 with
>>>> --dynamic-list-data -Bsymbolic and dlopen on libdl6a.so
>>>> W: [ld-elf/shared.exp] REGRESSION (PASS -> FAIL): Run dl6a5 with
>>>> -Bsymbolic-functions --dynamic-list-cpp-new and dlopen on libdl6a.so
>>>> W: [ld-elf/shared.exp] REGRESSION (PASS -> FAIL): Run dl6a2 with
>>>> -Bsymbolic-functions and dlopen on libdl6a.so
>>>> W: [ld-i386/i386.exp] REGRESSION (PASS -> FAIL): Run got1
>>>> W: [ld-i386/i386.exp] REGRESSION (PASS -> FAIL): Build gotpc1
>>>> W: [ld-i386/i386.exp] REGRESSION (PASS -> FAIL): Undefined weak symbol (-fPIE )
>>>> W: [ld-i386/i386.exp] REGRESSION (PASS -> FAIL): Undefined weak symbol (-fPIE -pie)
>>>> W: [ld-i386/i386.exp] REGRESSION (PASS -> FAIL): Run pr19031
>>>> W: [ld-srec/srec.exp] REGRESSION (PASS -> FAIL): S-records
>>>> W: [ld-srec/srec.exp] REGRESSION (PASS -> FAIL): S-records with constructors
>>>> W: [ld-gc/gc.exp] REGRESSION (PASS -> FAIL): --gc-sections with KEEP
>>>> W: [ld-size/size.exp] REGRESSION (PASS -> FAIL): Run size-8
>>>> W: [ld-i386/no-plt.exp] REGRESSION (PASS -> FAIL): No PLT (dynamic 1i)
>>>> W: [ld-i386/no-plt.exp] REGRESSION (PASS -> FAIL): Run pr20253-1f
>>>> W: [ld-i386/no-plt.exp] REGRESSION (PASS -> FAIL): Run pr20244-3a
>>>> W: [ld-i386/no-plt.exp] REGRESSION (PASS -> FAIL): Run pr20253-2a
>>>> W: [ld-i386/no-plt.exp] REGRESSION (PASS -> FAIL): No PLT (dynamic 1h)
>>>> W: [ld-i386/no-plt.exp] REGRESSION (PASS -> FAIL): Run pr20253-2c
>>>> 33 REGRESSIONS (1.18%).
>>>> 4 progressions (0.14%).
>>>> 2804 tests: 2750 pass (98.07%), 33 fail (1.18%), 4 xfail (0.14%) 1 untested (0.04%).
>>>>
>>>
>>> I can't reproduce it on x86-64.
>>>
>>
>> With PIE as default, I saw
>
> right, all compilers are configured/patched to default to hardening settings.
>

 x86-64 is clean now.

-- 
H.J.