Bug 68610 - lcms upgrade from 1.12,1 -> 1.13,1 fails: "Coarse error! 16 on entry 1021"
Summary: lcms upgrade from 1.12,1 -> 1.13,1 fails: "Coarse error! 16 on entry 1021"
Status: Closed FIXED
Alias: None
Product: Ports & Packages
Classification: Unclassified
Component: Individual Port(s) (show other bugs)
Version: Latest
Hardware: Any Any
: Normal Affects Only Me
Assignee: freebsd-ports-bugs (Nobody)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2004-07-02 20:00 UTC by david
Modified: 2004-10-26 00:14 UTC (History)
0 users

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description david 2004-07-02 20:00:38 UTC
After doing the daily build of -STABLE, I boot it up, then run

	portupgrade -a

(during the -STABLE build, I had already done a "cvs update" to the
ports tree).

(I do this within "script," in order to retain a record of what happened.)

Here is a cut/paste from the beginning, then the end of the portupgrade
run for lcms:

--->  Upgrading 'lcms-1.12,1' to 'lcms-1.13,1' (graphics/lcms)
--->  Building '/usr/ports/graphics/lcms'
===>  Cleaning for libtool-1.3.5_2
===>  Cleaning for jpeg-6b_3
===>  Cleaning for tiff-3.6.1_1
===>  Cleaning for lcms-1.13,1
===>  Vulnerability check disabled
===>  Extracting for lcms-1.13,1

Fix: 

Unkown, but I'll be happy to test possible approaches.
How-To-Repeat: 
"portupgrade -a" works for me; I suspect that "portupgrade lcms-1.12,1" would
be as effective.
Comment 1 Volker Stolz freebsd_committer freebsd_triage 2004-07-05 09:50:33 UTC
State Changed
From-To: open->feedback

Sorry, can't reproduce. Do you set any compiler-flags in /etc/make.conf? 


Comment 2 Volker Stolz freebsd_committer freebsd_triage 2004-07-05 09:50:33 UTC
Responsible Changed
From-To: freebsd-ports-bugs->vs

Sorry, can't reproduce. Do you set any compiler-flags in /etc/make.conf?
Comment 3 david 2004-07-06 17:59:59 UTC
I had been asked (apparently other than as a follow-up, since I don't
see the exchange in the PR) about non-default compiler options in
/etc/make.conf; I responded with the /etc/make.conf contents for each
of 4 systems that exhibit the problem.

As a further experiment, on my work desktop (2.4 GHz P4, UP), I moved
/etc/make.conf aside, then tried "portupgrade lcms-1.12,1".

Once again, the failure exhibits itself:

frecnocpc2(4.10-S)[15] mv /etc/make.conf{,.save} && portupgrade lcms-1.12,1
--->  Upgrading 'lcms-1.12,1' to 'lcms-1.13,1' (graphics/lcms)
--->  Building '/usr/ports/graphics/lcms'
===>  Cleaning for libtool-1.3.5_2
===>  Cleaning for jpeg-6b_3
===>  Cleaning for tiff-3.6.1_1
===>  Cleaning for lcms-1.13,1
===>  Vulnerability check disabled
===>  Extracting for lcms-1.13,1
>> Checksum OK for lcms-1.13.tar.gz.
===>  Patching for lcms-1.13,1
===>  Applying FreeBSD patches for lcms-1.13,1
===>   lcms-1.13,1 depends on file: /usr/local/bin/libtool13 - found
===>   lcms-1.13,1 depends on shared library: tiff.4 - found
===>  Configuring for lcms-1.13,1
checking build system type... i386-portbld-freebsd4.10
checking host system type... i386-portbld-freebsd4.10
....
make  testcms
source='testcms.c' object='testcms.o' libtool=no  depfile='.deps/testcms.Po' tmpdepfile='.deps/testcms.TPo'  depmode=gcc /bin/sh ../depcomp  cc -DPACKAGE_NAME=\"lcms\" -DPACKAGE_TARNAME=\"lcms\" -DPACKAGE_VERSION=\"1.13\" -DPACKAGE_STRING=\"lcms\ 1.13\" -DPACKAGE_BUGREPORT=\"\" -DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_STAT_H=1 -DHAVE_STDLIB_H=1 -DHAVE_STRING_H=1 -DHAVE_MEMORY_H=1 -DHAVE_STRINGS_H=1 -DHAVE_INTTYPES_H=1 -DHAVE_UNISTD_H=1 -DHAVE_DLFCN_H=1 -DSIZEOF_INT=4 -DSIZEOF_UNSIGNED_INT=4 -DSIZEO F_LONG=4 -DSIZEOF_UNSIGNED_LONG=4 -DSIZEOF_LONG_LONG=8 -DSIZEOF_UNSIGNED_LONG_LONG=8 -DHasZLIB=1 -I. -I. -I../include -I../include   -I/usr/local/include  -O -pipe -c `test -f 'testcms.c' || echo './'`testcms.c
/bin/sh ../libtool --mode=link cc  -O -pipe   -L/usr/local/lib -o testcms -L/usr/local/lib testcms.o ../src/liblcms.la 
mkdir .libs
cc -O -pipe -L/usr/local/lib -o .libs/testcms -L/usr/local/lib testcms.o ../src/.libs/liblcms.so -lm  -Wl,--rpath -Wl,/usr/local/lib
creating testcms
./testcms
little cms testbed. Ver 1.13 [build Jul  6 2004 09:49:07]

Testing fixed point:    2.884667 = 2.884672
Testing fixed scaling...pass.
Testing curves join ...pass.
Testing reversing of curves ...Coarse error! 16 on entry 1021: FFD4/FFBE*** Error code 1

Stop in /common/ports/graphics/lcms/work/lcms-1.13/testbed.
*** Error code 1

Stop in /common/ports/graphics/lcms.
** Command failed [exit code 1]: /usr/bin/script -qa /tmp/portupgrade91463.0 make
** Fix the problem and try again.
** Listing the failed packages (*:skipped / !:failed)
        ! graphics/lcms (lcms-1.12,1)   (unknown build error)
--->  Packages processed: 0 done, 0 ignored, 0 skipped and 1 failed
frecnocpc2(4.10-S)[16] 


For reference, this system is:

frecnocpc2(4.10-S)[16] uname -a
FreeBSD frecnocpc2.noc.egation.com 4.10-STABLE FreeBSD 4.10-STABLE #19: Fri Jul  2 12:07:25 PDT 2004     root@frecnocpc2.noc.egation.com:/common/S1/obj/usr/src/sys/REPO  i386
frecnocpc2(4.10-S)[17] 

At this point, I'd be interested in seeing any system that builds the
port correctly....

Peace,
david
-- 
David H. Wolfskill				david@catwhisker.org
I do not "unsubscribe" from email "services" to which I have not explicitly
subscribed.  Rather, I block spammers' access to SMTP servers I control,
and encourage others who are in a position to do so to do likewise.
Comment 4 Volker Stolz freebsd_committer freebsd_triage 2004-07-09 10:00:46 UTC
It works fine here and on the cluster:

little cms testbed. Ver 1.13 [build Jul  1 2004 04:20:04]

Testing fixed point:    2.884667 = 2.884672
Testing fixed scaling...pass.
Testing curves join ...pass.
Testing reversing of curves ...
dE: mean=0.00123875, SD=0.00346281, max=0.0622568  pass.

http://pointyhat.freebsd.org/errorlogs/i386-4-latest-logs/lcms-1.13,1.log.bz2
Comment 5 david 2004-07-09 16:17:05 UTC
>To: freebsd-gnats-submit@freebsd.org, david@catwhisker.org
>Subject: Re: ports/68610: lcms upgrade from 1.12,1 -> 1.13,1 fails: "Coarse error! 16 on entry 1021"
>From: "Volker Stolz" <vs@freebsd.org>
>Date: Fri, 09 Jul 2004 11:00:46 +0200

>It works fine here and on the cluster:

OK... so:  what can I(/we) do to figure out what's going wrong with
every system I try?  At this point, I have no idea what a "coarse error"
might be, let alone what would cause one to be detected; where would I
get more clues (ideally, without turning this into a career unto
itself)?

Thanks,
david
-- 
David H. Wolfskill				david@catwhisker.org
I do not "unsubscribe" from email "services" to which I have not explicitly
subscribed.  Rather, I block spammers' access to SMTP servers I control,
and encourage others who are in a position to do so to do likewise.
Comment 6 Volker Stolz freebsd_committer freebsd_triage 2004-07-23 14:31:55 UTC
State Changed
From-To: feedback->suspended

Suspend and return to pool. I'm really out of wits here, sorry. 
I'm glad the package fixed it, though. Maybe you should contact 
the lcms-authors. 


Comment 7 Volker Stolz freebsd_committer freebsd_triage 2004-07-23 14:31:55 UTC
Responsible Changed
From-To: vs->freebsd-ports-bugs

Suspend and return to pool. I'm really out of wits here, sorry. 
I'm glad the package fixed it, though. Maybe you should contact 
the lcms-authors.
Comment 8 M.Handley 2004-09-11 16:29:15 UTC
I just hit the same problem.  As far as I can see, it looks like the
problem is that testcms (the inbuilt test suite) gets its header files
from the installed lcms rather than the one in its source directory,
and the two are not compatible.  Obviously this doesn't happen with a
fresh install, but does happen when using portupgrade.

A simple crude workaround is to pkg_delete -f <oldliblcms>, before
building and installing the new one.

 - Mark
Comment 9 stolz 2004-09-13 12:59:26 UTC
In gmane.os.freebsd.devel.ports.bugs, you wrote:
>  I just hit the same problem.  As far as I can see, it looks like the
>  problem is that testcms (the inbuilt test suite) gets its header files
>  from the installed lcms rather than the one in its source directory,
>  and the two are not compatible.  Obviously this doesn't happen with a
>  fresh install, but does happen when using portupgrade.

Hm, looking at the compiler-invocation, I don't see any potential problem:
(apart maybe from libtool not DTRT):

make  testcms
source='testcms.c' object='testcms.o' libtool=no  depfile='.deps/testcms.Po' tmpdepfile='.deps/testcms.TPo'  depmode=gcc /bin/sh ../depcomp  cc -DPACKAGE_NAME=\"lcms\" -DPACKAGE_TARNAME=\"lcms\" -DPACKAGE_VERSION=\"1.13\" -DPACKAGE_STRING=\"lcms\ 1.13\" -DPACKAGE_BUGREPORT=\"\" -DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_STAT_H=1 -DHAVE_STDLIB_H=1 -DHAVE_STRING_H=1 -DHAVE_MEMORY_H=1 -DHAVE_STRINGS_H=1 -DHAVE_INTTYPES_H=1 -DHAVE_UNISTD_H=1 -DHAVE_DLFCN_H=1 -DSIZEOF_INT=4 -DSIZEOF_UNSIGNED_INT=4 -DSIZEO F_LONG=4 -DSIZEOF_UNSIGNED_LONG=4 -DSIZEOF_LONG_LONG=8 -DSIZEOF_UNSIGNED_LONG_LONG=8 -DHasZLIB=1 -I. -I. -I../include -I../include   -I/usr/local/include  -O -pipe -g -c `test -f 'testcms.c' || echo './'`testcms.c
/bin/sh ../libtool --mode=link cc  -O -pipe -g   -L/usr/local/lib -o testcms -L/usr/local/lib testcms.o ../src/liblcms.la
mkdir .libs
cc -O -pipe -g -L/usr/local/lib -o .libs/testcms -L/usr/local/lib testcms.o ../src/.libs/liblcms.so -lm  -Wl,--rpath -Wl,/usr/local/lib

I'll see if I can reproduce this here.
-- 
http://www-i2.informatik.rwth-aachen.de/stolz/ *** PGP *** S/MIME
Two more month.
Comment 10 M.Handley 2004-09-13 13:29:05 UTC
>Hm, looking at the compiler-invocation, I don't see any potential problem:
>(apart maybe from libtool not DTRT):

Unfortunately I didn't keep good enough notes as to exactly what
happened.  Here's what I do know:

I was upgrading from 1.09 to 1.13, and hit this problem with testcms.
Then I tried upgrading from 1.09 to 1.12, and still hit the problem,
but fewer tests failed.  So then I instrumented the library and test
suite, and discovered that some of the internal tables were being
initialized to all zeros instead of having the correct data.  Usually
this sort of thing only happens when an API has changed.  So then I
forced a pkg_delete on lcms (no other change), and rebuilt 1.13.  The
internal tests then all passed.

So, I don't know for sure that this is a header file issue (that was
my most likely hypothesis) or a linkage issue, but it is certainly a
bad interaction between the test suite "testcms" and the previously
installed library.

Cheers,
	Mark
Comment 11 Mikhail Teterin freebsd_committer freebsd_triage 2004-10-26 00:12:11 UTC
State Changed
From-To: suspended->closed

Today's version of the port links the liblcms into the testcms utility 
statically. This should fix the upgrade problems.