Mirror of the gdb mailing list
 help / color / mirror / Atom feed
* prelink -u on aarch64 question
@ 2017-07-26 17:37 Steve Ellcey
  2017-07-26 19:44 ` Jan Kratochvil
  0 siblings, 1 reply; 3+ messages in thread
From: Steve Ellcey @ 2017-07-26 17:37 UTC (permalink / raw)
  To: gdb

I was wondering if anyone could tell me anything about prelink on
aarch64? I am looking at a gdb test run on my aarch64 ubuntu 16.04
system using the installed gcc (5.4.0) and prelink (1.0) and top-of-tree
gdb sources.  When I run the gdb testsuite and compare my results
with ones from the gdb testing mailing list I get several failures
that involve prelink and the -u option.  It looks like the test run
from the mailing list has XFAIL's for tests that use prelink because
it is not installed on the system.  I get more FAIL's when I remove
prelink, not XFAIL's, I am not sure why.  On my system when I use prelink
with the -u option it gives this message:

+ /usr/sbin/prelink -uN --dynamic-linker=/home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread.d/ld-linux-aarch64.so.1 --ld-library-path=/home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread.d /home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread /home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread.d/libm.so.6 /home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread.d/libc.so.6 /home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread.d/ld-linux-aarch64.so.1
prelink: "/home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread"'s architecture is not supported
prelink: "/home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread.d/libm.so.6"'s architecture is not supported
prelink: "/home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread.d/libc.so.6"'s architecture is not supported
prelink: "/home/sellcey/gdb-std/obj/binutils-gdb/gdb/testsuite/outputs/gdb.base/attach-pie-misread/attach-pie-misread.d/ld-linux-aarch64.so.1"'s architecture is not supported


Steve Ellcey
sellcey@cavium.com


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: prelink -u on aarch64 question
  2017-07-26 17:37 prelink -u on aarch64 question Steve Ellcey
@ 2017-07-26 19:44 ` Jan Kratochvil
  2017-07-26 20:05   ` Steve Ellcey
  0 siblings, 1 reply; 3+ messages in thread
From: Jan Kratochvil @ 2017-07-26 19:44 UTC (permalink / raw)
  To: Steve Ellcey; +Cc: gdb

On Wed, 26 Jul 2017 19:37:46 +0200, Steve Ellcey wrote:
> I was wondering if anyone could tell me anything about prelink on
> aarch64?

prelink probably is not supported on aarch64 at all, I have not checked.

But even on x86* prelink has been discontinued - on its "core" distro:
	Don't enable prelink by default in Fedora
	https://pagure.io/fesco/issue/1183

I think at least the prelink testfile gdb.base/break-interp.exp could be even
disabled by default nowadays (saying that as author of the testfile) as even
when it PASSes it is very slow given the huge number of combinations it does
test.


Jan


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: prelink -u on aarch64 question
  2017-07-26 19:44 ` Jan Kratochvil
@ 2017-07-26 20:05   ` Steve Ellcey
  0 siblings, 0 replies; 3+ messages in thread
From: Steve Ellcey @ 2017-07-26 20:05 UTC (permalink / raw)
  To: Jan Kratochvil; +Cc: gdb

On Wed, 2017-07-26 at 21:44 +0200, Jan Kratochvil wrote:
> On Wed, 26 Jul 2017 19:37:46 +0200, Steve Ellcey wrote:
> > 
> > I was wondering if anyone could tell me anything about prelink on
> > aarch64?
> prelink probably is not supported on aarch64 at all, I have not checked.
> 
> But even on x86* prelink has been discontinued - on its "core" distro:
> 	Don't enable prelink by default in Fedora
> 	https://pagure.io/fesco/issue/1183
> 
> I think at least the prelink testfile gdb.base/break-interp.exp could be even
> disabled by default nowadays (saying that as author of the testfile) as even
> when it PASSes it is very slow given the huge number of combinations
> it does
> test.
> 
> 
> Jan

I don't see any indication that gdb.base/break-interp.exp is disabled
by default, though it is part of the slow_tests list so if you don't
run slow tests then it would not show up.  I actually have four tests
that fail due to prelink:

gdb.base/attach-pie-misread.exp
gdb.base/break-interp.exp
gdb.base/prelink.exp
gdb.threads/dlopen-libpthread.exp

It also seems to get used in:

gdb.server/connect-stopped-target.exp
gdb.server/solib-list.exp
gdb.base/nostdlib.exp
gdb.base/relativedebug.exp
gdb.base/solib-overlap.exp

Although those did not show up in my FAIL list.

Should we just remove all the gdb tests that use prelink?

Steve Ellcey
sellcey@cavium.com


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2017-07-26 20:05 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-07-26 17:37 prelink -u on aarch64 question Steve Ellcey
2017-07-26 19:44 ` Jan Kratochvil
2017-07-26 20:05   ` Steve Ellcey

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox