Debian Bug report logs - #432658
monotone: FTBFS: internal compiler error on arm in database_check.cc

version graph

Package: monotone; Maintainer for monotone is Debian Maintainers for Monotone <monotone-debian@nongnu.org>; Source for monotone is src:monotone.

Reported by: Ludovic Brenta <ludovic@ludovic-brenta.org>

Date: Wed, 11 Jul 2007 08:39:02 UTC

Severity: serious

Found in version monotone/0.35-1

Fixed in version monotone/0.36-1

Done: "Zack Weinberg" <zackw@panix.com>

Bug is archived. No further changes may be made.

Toggle useless messages

View this report as an mbox folder, status mbox, maintainer mbox


Report forwarded to debian-bugs-dist@lists.debian.org, Shaun Jackman <sjackman@debian.org>:
Bug#432658; Package monotone. Full text and rfc822 format available.

Acknowledgement sent to Ludovic Brenta <ludovic@ludovic-brenta.org>:
New Bug report received and forwarded. Copy sent to Shaun Jackman <sjackman@debian.org>. Full text and rfc822 format available.

Message #5 received at submit@bugs.debian.org (full text, mbox):

From: Ludovic Brenta <ludovic@ludovic-brenta.org>
To: submit@bugs.debian.org
Subject: monotone: FTBFS: internal compiler error on arm in database_check.cc
Date: Wed, 11 Jul 2007 10:37:25 +0200
Package: monotone
Version: 0.35-1
Severity: serious

The buildd log on arm says:

g++ -DLOCALEDIR=\"/usr/share/locale\" -DHAVE_CONFIG_H -I. -I. -I.  -I./lua -I./sqlite   -DNDEBUG -DBOOST_DISABLE_THREADS -DBOOST_MULTI_INDEX_DISABLE_SERIALIZATION   -g -Wall -O2 -Wall -W -Wno-unused -fno-strict-aliasing -c -o mtn-database_check.o `test -f 'database_check.cc' || echo './'`database_check.cc
database_check.cc: In function 'void report_certs(const std::map<revision<hexenc<id> >, checked_revision, std::less<revision<hexenc<id> > >, std::allocator<std::pair<const revision<hexenc<id> >, checked_revision> > >&, size_t&, size_t&, size_t&, size_t&)':
database_check.cc:835: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.

I am well aware that this is a compiler bug but before submitting it against
g++-4.1, I would like to see if we can find a workaround in monotone.
Failing that, I suggest we drop arm from the list of supported architectures
for this package.

BTW, it is possible that the compiler bug is already known.  I haven't
looked in the GC bug database yet.

-- 
Ludovic Brenta.



Information forwarded to debian-bugs-dist@lists.debian.org, Shaun Jackman <sjackman@debian.org>:
Bug#432658; Package monotone. Full text and rfc822 format available.

Acknowledgement sent to Sune Vuorela <debian@pusling.com>:
Extra info received and forwarded to list. Copy sent to Shaun Jackman <sjackman@debian.org>. Full text and rfc822 format available.

Message #10 received at 432658@bugs.debian.org (full text, mbox):

From: Sune Vuorela <debian@pusling.com>
To: Ludovic Brenta <ludovic@ludovic-brenta.org>, 432658@bugs.debian.org
Subject: Re: Bug#432658: monotone: FTBFS: internal compiler error on arm in database_check.cc
Date: Wed, 11 Jul 2007 10:49:41 +0200
On Wednesday 11 July 2007, Ludovic Brenta wrote:
> Package: monotone
> Version: 0.35-1
> Severity: serious
>
> The buildd log on arm says:
>
> g++ -DLOCALEDIR=\"/usr/share/locale\" -DHAVE_CONFIG_H -I. -I. -I.  -I./lua
> -I./sqlite   -DNDEBUG -DBOOST_DISABLE_THREADS
> -DBOOST_MULTI_INDEX_DISABLE_SERIALIZATION   -g -Wall -O2 -Wall -W
> -Wno-unused -fno-strict-aliasing -c -o mtn-database_check.o `test -f
> 'database_check.cc' || echo './'`database_check.cc database_check.cc: In
> function 'void report_certs(const std::map<revision<hexenc<id> >,
> checked_revision, std::less<revision<hexenc<id> > >,
> std::allocator<std::pair<const revision<hexenc<id> >, checked_revision> >
> >&, size_t&, size_t&, size_t&, size_t&)': database_check.cc:835: internal
> compiler error: Segmentation fault Please submit a full bug report,
> with preprocessed source if appropriate.
>
> I am well aware that this is a compiler bug but before submitting it
> against g++-4.1, I would like to see if we can find a workaround in
> monotone. Failing that, I suggest we drop arm from the list of supported
> architectures for this package.


This might also be because it has been tried to be built on one of the slower 
arm buildds - and the cc1plus command have used too much memory - and then 
gets killed.

I have at least seen that in other packages.

/Sune
 - who doesn't know anything about monotone
-- 
Man, how to receive the device?

You either should never debug the level-9 monitor of a window, or never have 
to ping to a IRC hard disk.



Information forwarded to debian-bugs-dist@lists.debian.org, Debian Maintainers for Monotone <monotone-debian@nongnu.org>:
Bug#432658; Package monotone. Full text and rfc822 format available.

Acknowledgement sent to Ludovic Brenta <ludovic@ludovic-brenta.org>:
Extra info received and forwarded to list. Copy sent to Debian Maintainers for Monotone <monotone-debian@nongnu.org>. Full text and rfc822 format available.

Message #15 received at 432658@bugs.debian.org (full text, mbox):

From: Ludovic Brenta <ludovic@ludovic-brenta.org>
To: 432658@bugs.debian.org
Cc: arm@buildd.debian.org
Subject: Re: #432658: monotone: FTBFS: internal compiler error on arm in database_check.cc
Date: Sun, 15 Jul 2007 15:44:43 +0200
I crudely measured, using top(1) and a sampling period of 0.5 s, that
g++ uses as much as 474 Mb of virtual memory when compiling the
largest file in monotone (revision.cc).  That might explain why g++
seems to have run out of memory on two different arm machines,
smackdown and cats.

-- 
Ludovic Brenta.




Information forwarded to debian-bugs-dist@lists.debian.org, Debian Maintainers for Monotone <monotone-debian@nongnu.org>:
Bug#432658; Package monotone. Full text and rfc822 format available.

Acknowledgement sent to "Zack Weinberg" <zackw@panix.com>:
Extra info received and forwarded to list. Copy sent to Debian Maintainers for Monotone <monotone-debian@nongnu.org>. Full text and rfc822 format available.

Message #20 received at 432658@bugs.debian.org (full text, mbox):

From: "Zack Weinberg" <zackw@panix.com>
To: "Ludovic Brenta" <ludovic@ludovic-brenta.org>, 432658@bugs.debian.org
Cc: arm@buildd.debian.org
Subject: Re: [Monotone-debian] Bug#432658: #432658: monotone: FTBFS: internal compiler error on arm in database_check.cc
Date: Sun, 15 Jul 2007 17:04:57 -0700
On 7/15/07, Ludovic Brenta <ludovic@ludovic-brenta.org> wrote:
> I crudely measured, using top(1) and a sampling period of 0.5 s, that
> g++ uses as much as 474 Mb of virtual memory when compiling the
> largest file in monotone (revision.cc).  That might explain why g++
> seems to have run out of memory on two different arm machines,
> smackdown and cats.

I don't think this is out of memory, I think this is a real compiler
bug.  The crash is in the same place for both failed builds:

database_check.cc: In function 'void report_certs(const
std::map<revision<hexenc<id> >, checked_revision,
std::less<revision<hexenc<id> > >, std::allocator<std::pair<const
revision<hexenc<id> >, checked_revision> > >&, size_t&, size_t&,
size_t&, size_t&)':
database_check.cc:835: internal compiler error: Segmentation fault

The function in question is not terribly complicated - and compilation
of several more complicated functions/files has already succeeded at
the point where the buildd failed.

zw



Information forwarded to debian-bugs-dist@lists.debian.org, Debian Maintainers for Monotone <monotone-debian@nongnu.org>:
Bug#432658; Package monotone. Full text and rfc822 format available.

Acknowledgement sent to "Zack Weinberg" <zackw@panix.com>:
Extra info received and forwarded to list. Copy sent to Debian Maintainers for Monotone <monotone-debian@nongnu.org>. Full text and rfc822 format available.

Message #25 received at 432658@bugs.debian.org (full text, mbox):

From: "Zack Weinberg" <zackw@panix.com>
To: 432658@bugs.debian.org
Subject: plan
Date: Sat, 21 Jul 2007 18:12:44 -0700
For the record - since I'm pretty sure this is a compiler bug and the
g++ 4.2 transition is imminent, I'm going to wait and see if the
package builds with 4.2, and if it does, I'll downgrade this to normal
and reassign it to g++-4.1.

zw



Information forwarded to debian-bugs-dist@lists.debian.org, Debian Maintainers for Monotone <monotone-debian@nongnu.org>:
Bug#432658; Package monotone. Full text and rfc822 format available.

Acknowledgement sent to Ludovic Brenta <ludovic@ludovic-brenta.org>:
Extra info received and forwarded to list. Copy sent to Debian Maintainers for Monotone <monotone-debian@nongnu.org>. Full text and rfc822 format available.

Message #30 received at 432658@bugs.debian.org (full text, mbox):

From: Ludovic Brenta <ludovic@ludovic-brenta.org>
To: "Zack Weinberg" <zackw@panix.com>
Cc: 432658@bugs.debian.org
Subject: Re: Bug#432658: plan
Date: Mon, 23 Jul 2007 00:59:51 +0200
Not so imminent.  There is some pushback from the linux and Mozilla
maintainers because gcc-4.2 introduces build failures in those
packages (see the archives of debian-release for the past couple of
days).  Also, the glibc maintainers would like to stage a transition
to a new version of glibc before the transition to g++-4.2.

Monotone 0.35-2 is now 8 days old, but I very much doubt it will enter
testing on Tuesday or Wednesday, because it is waiting for boost,
which itself has RC bugs preventing it from entering testing.

Therefore, I don't think monotone is under any kind of pressure.  I
would rather suggest to move forward with the long-term plan to not
depend on boost anymore, and upload 0.36 whenever it is ready.  If g++
is still a problem at that point, let's just disable arm temporarily,
and reinstate it later when g++ is fixed.

-- 
Ludovic Brenta.




Information forwarded to debian-bugs-dist@lists.debian.org, Debian Maintainers for Monotone <monotone-debian@nongnu.org>:
Bug#432658; Package monotone. Full text and rfc822 format available.

Acknowledgement sent to "Zack Weinberg" <zackw@panix.com>:
Extra info received and forwarded to list. Copy sent to Debian Maintainers for Monotone <monotone-debian@nongnu.org>. Full text and rfc822 format available.

Message #35 received at 432658@bugs.debian.org (full text, mbox):

From: "Zack Weinberg" <zackw@panix.com>
To: "Ludovic Brenta" <ludovic@ludovic-brenta.org>
Cc: 432658@bugs.debian.org
Subject: Re: Bug#432658: plan
Date: Sun, 22 Jul 2007 17:47:18 -0700
On 7/22/07, Ludovic Brenta <ludovic@ludovic-brenta.org> wrote:
> Not so imminent.  [...]

Imminent was perhaps the wrong word, but nonetheless, I don't see much
value in trying to isolate a bug in g++-4.1 when 4.2 should be live in
the near future, especially when, as you say

> Monotone 0.35-2 is now 8 days old, but I very much doubt it will enter
> testing on Tuesday or Wednesday, because it is waiting for boost,
> which itself has RC bugs preventing it from entering testing.
> Therefore, I don't think monotone is under any kind of pressure.

which is very much what I was thinking myself.

> would rather suggest to move forward with the long-term plan to not
> depend on boost anymore, and upload 0.36 whenever it is ready.  If g++
> is still a problem at that point, let's just disable arm temporarily,
> and reinstate it later when g++ is fixed.

0.36 should be ready in a few days; upstream proceeds with boost
removal, but it won't happen for that release.  More likely 0.37.  My
inclination is to leave Architecture: alone for the 0.36 release --
perhaps it will miraculously start working on arm, after all.

zw



Information forwarded to debian-bugs-dist@lists.debian.org, Debian Maintainers for Monotone <monotone-debian@nongnu.org>:
Bug#432658; Package monotone. Full text and rfc822 format available.

Acknowledgement sent to Ludovic Brenta <ludovic@ludovic-brenta.org>:
Extra info received and forwarded to list. Copy sent to Debian Maintainers for Monotone <monotone-debian@nongnu.org>. Full text and rfc822 format available.

Message #40 received at 432658@bugs.debian.org (full text, mbox):

From: Ludovic Brenta <ludovic@ludovic-brenta.org>
To: "Zack Weinberg" <zackw@panix.com>
Cc: 432658@bugs.debian.org, "Ludovic Brenta" <ludovic@ludovic-brenta.org>
Subject: Re: Bug#432658: plan
Date: Mon, 23 Jul 2007 20:13:57 +0200
Zack Weinberg writes:
> 0.36 should be ready in a few days; upstream proceeds with boost
> removal, but it won't happen for that release.  More likely 0.37.  My
> inclination is to leave Architecture: alone for the 0.36 release --
> perhaps it will miraculously start working on arm, after all.

Agreed.  Since 0.36 will still depend on boost, there is no point in
disabling arm, either, since boost will still block it from testing.

As we both said: no pressure :)

-- 
Ludovic Brenta.




Reply sent to "Zack Weinberg" <zackw@panix.com>:
You have taken responsibility. Full text and rfc822 format available.

Notification sent to Ludovic Brenta <ludovic@ludovic-brenta.org>:
Bug acknowledged by developer. Full text and rfc822 format available.

Message #45 received at 432658-done@bugs.debian.org (full text, mbox):

From: "Zack Weinberg" <zackw@panix.com>
To: 432658-done@bugs.debian.org
Subject: fixed in 0.36-1
Date: Sun, 26 Aug 2007 11:05:02 -0700
Package: monotone
Version: 0.36-1

The 0.36-1 package was built successfully for arm
(<http://buildd.debian.org/fetch.cgi?pkg=monotone;ver=0.36-1;arch=arm;stamp=1188050397>)
so closing this bug.

zw



Bug archived. Request was from Debbugs Internal Request <owner@bugs.debian.org> to internal_control@bugs.debian.org. (Wed, 09 Apr 2008 07:32:14 GMT) Full text and rfc822 format available.

Send a report that this bug log contains spam.


Debian bug tracking system administrator <owner@bugs.debian.org>. Last modified: Sun Apr 20 21:04:47 2014; Machine Name: buxtehude.debian.org

Debian Bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.