linux-m68k.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* strip is running out of memory on m68k (Haskell)
@ 2018-07-31 12:42 John Paul Adrian Glaubitz
  2018-08-01  8:37 ` John Paul Adrian Glaubitz
  0 siblings, 1 reply; 6+ messages in thread
From: John Paul Adrian Glaubitz @ 2018-07-31 12:42 UTC (permalink / raw)
  To: Debian m68k; +Cc: Linux/m68k

Hi!

We have recently made lots of progress fixing bugs on m68k both in various
as well as QEMU. This has lead to the list of successfully built packages
to over 11.500 - Debian/m68k never built so many packages before.

Now, there are a few things that still need to work and one of them is
binutils' strip command which is bailing out for a very low number of
Haskell packages [1]:

debian/hlibrary.setup copy --builddir=dist-ghc --destdir=debian/tmp-inst-ghc
Installing library in debian/tmp-inst-ghc/usr/lib/haskell-packages/ghc/lib/m68k-linux-ghc-8.2.2/criterion-1.3.0.0-CVIsDHnXrrBAzasYPdPd4j
Installing executable criterion-report in debian/tmp-inst-ghc/usr/bin
Warning: The directory debian/tmp-inst-ghc/usr/bin is not in the system search
path.
/usr/bin/strip:debian/tmp-inst-ghc/usr/bin/criterion-report: memory exhausted
make: *** [/usr/share/cdbs/1/class/hlibrary.mk:188: debian/tmp-inst-ghc] Error 1
dpkg-buildpackage: error: fakeroot debian/rules binary-arch subprocess returned exit status 2

Since I don't think that we have any actual memory limitations on qemu-user (as
compared to qemu-system), this issue might be a result of strip having a hardwired
maximum buffer size for stripping binaries.

Does anyone know more? If yes, could we patch strip to allow larger binaries?

Adrian

> [1] https://buildd.debian.org/status/fetch.php?pkg=haskell-criterion&arch=m68k&ver=1.3.0.0-1&stamp=1533028852&raw=0

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: strip is running out of memory on m68k (Haskell)
  2018-07-31 12:42 strip is running out of memory on m68k (Haskell) John Paul Adrian Glaubitz
@ 2018-08-01  8:37 ` John Paul Adrian Glaubitz
  2018-08-01  9:02   ` John Paul Adrian Glaubitz
  0 siblings, 1 reply; 6+ messages in thread
From: John Paul Adrian Glaubitz @ 2018-08-01  8:37 UTC (permalink / raw)
  To: Debian m68k; +Cc: Linux/m68k

On 07/31/2018 02:42 PM, John Paul Adrian Glaubitz wrote:
> Since I don't think that we have any actual memory limitations on qemu-user (as
> compared to qemu-system), this issue might be a result of strip having a hardwired
> maximum buffer size for stripping binaries.
> 
> Does anyone know more? If yes, could we patch strip to allow larger binaries?

Ok, this seems to be a regression. Packages built with binutils_2.31.1-1 don't
show the problem while those built with 2.31.1-2 do:

With binutils_2.31.1-1:

> https://buildd.debian.org/status/fetch.php?pkg=wireshark&arch=m68k&ver=2.6.2-1&stamp=1532706397&raw=0

With binutils_2.31.1-2:

> https://buildd.debian.org/status/fetch.php?pkg=wireshark&arch=m68k&ver=2.6.2-2&stamp=1533058050&raw=0

Will write a bug report upstream once I can confirm this.

Adrian

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: strip is running out of memory on m68k (Haskell)
  2018-08-01  8:37 ` John Paul Adrian Glaubitz
@ 2018-08-01  9:02   ` John Paul Adrian Glaubitz
  2018-08-01  9:58     ` John Paul Adrian Glaubitz
  0 siblings, 1 reply; 6+ messages in thread
From: John Paul Adrian Glaubitz @ 2018-08-01  9:02 UTC (permalink / raw)
  To: Debian m68k; +Cc: Linux/m68k

On 08/01/2018 10:37 AM, John Paul Adrian Glaubitz wrote:
> With binutils_2.31.1-1:
> 
>> https://buildd.debian.org/status/fetch.php?pkg=wireshark&arch=m68k&ver=2.6.2-1&stamp=1532706397&raw=0
> 
> With binutils_2.31.1-2:
> 
>> https://buildd.debian.org/status/fetch.php?pkg=wireshark&arch=m68k&ver=2.6.2-2&stamp=1533058050&raw=0
> 
> Will write a bug report upstream once I can confirm this.

Ok, downgrading to 2.31.1-1 definitely helps:

(sid-m68k-sbuild)root@epyc:/build/haskell-yaml-bukQp5/haskell-yaml-0.8.31.1/debian# strip tmp-inst-ghc/usr/bin/yaml2json
/usr/bin/strip:tmp-inst-ghc/usr/bin/yaml2json: memory exhausted
(sid-m68k-sbuild)root@epyc:/build/haskell-yaml-bukQp5/haskell-yaml-0.8.31.1/debian# strip tmp-inst-ghc/usr/bin/yaml2json
(sid-m68k-sbuild)root@epyc:/build/haskell-yaml-bukQp5/haskell-yaml-0.8.31.1/debian#

Adrian

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: strip is running out of memory on m68k (Haskell)
  2018-08-01  9:02   ` John Paul Adrian Glaubitz
@ 2018-08-01  9:58     ` John Paul Adrian Glaubitz
  2018-08-01 14:40       ` John Paul Adrian Glaubitz
  0 siblings, 1 reply; 6+ messages in thread
From: John Paul Adrian Glaubitz @ 2018-08-01  9:58 UTC (permalink / raw)
  To: Debian m68k; +Cc: Linux/m68k

On 08/01/2018 11:02 AM, John Paul Adrian Glaubitz wrote:
> Ok, downgrading to 2.31.1-1 definitely helps:
> 
> (sid-m68k-sbuild)root@epyc:/build/haskell-yaml-bukQp5/haskell-yaml-0.8.31.1/debian# strip tmp-inst-ghc/usr/bin/yaml2json
> /usr/bin/strip:tmp-inst-ghc/usr/bin/yaml2json: memory exhausted
> (sid-m68k-sbuild)root@epyc:/build/haskell-yaml-bukQp5/haskell-yaml-0.8.31.1/debian# strip tmp-inst-ghc/usr/bin/yaml2json
> (sid-m68k-sbuild)root@epyc:/build/haskell-yaml-bukQp5/haskell-yaml-0.8.31.1/debian#

I have filed https://sourceware.org/bugzilla/show_bug.cgi?id=23471 now.

Trying to find the commit now which introduced the regression.

Adrian

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: strip is running out of memory on m68k (Haskell)
  2018-08-01  9:58     ` John Paul Adrian Glaubitz
@ 2018-08-01 14:40       ` John Paul Adrian Glaubitz
  2018-08-02 14:04         ` Wouter Verhelst
  0 siblings, 1 reply; 6+ messages in thread
From: John Paul Adrian Glaubitz @ 2018-08-01 14:40 UTC (permalink / raw)
  To: Debian m68k; +Cc: Linux/m68k

On 08/01/2018 11:58 AM, John Paul Adrian Glaubitz wrote:
> I have filed https://sourceware.org/bugzilla/show_bug.cgi?id=23471 now.
> 
> Trying to find the commit now which introduced the regression.

After some debugging, it turns out to be a regression with gcc-8.

Building binutils with gcc-7 resolves the problem.

Adrian

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: strip is running out of memory on m68k (Haskell)
  2018-08-01 14:40       ` John Paul Adrian Glaubitz
@ 2018-08-02 14:04         ` Wouter Verhelst
  0 siblings, 0 replies; 6+ messages in thread
From: Wouter Verhelst @ 2018-08-02 14:04 UTC (permalink / raw)
  To: John Paul Adrian Glaubitz; +Cc: Debian m68k, Linux/m68k

On Wed, Aug 01, 2018 at 04:40:30PM +0200, John Paul Adrian Glaubitz wrote:
> On 08/01/2018 11:58 AM, John Paul Adrian Glaubitz wrote:
> > I have filed https://sourceware.org/bugzilla/show_bug.cgi?id=23471 now.
> > 
> > Trying to find the commit now which introduced the regression.
> 
> After some debugging, it turns out to be a regression with gcc-8.
> 
> Building binutils with gcc-7 resolves the problem.

You're reminding me of the fun I had with #327780 ;-)

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2018-08-02 14:04 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-07-31 12:42 strip is running out of memory on m68k (Haskell) John Paul Adrian Glaubitz
2018-08-01  8:37 ` John Paul Adrian Glaubitz
2018-08-01  9:02   ` John Paul Adrian Glaubitz
2018-08-01  9:58     ` John Paul Adrian Glaubitz
2018-08-01 14:40       ` John Paul Adrian Glaubitz
2018-08-02 14:04         ` Wouter Verhelst

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).