MapServer incompatibility with RHEL4/CE NTOS4.2

[Doyon, Jean-Francois]

Alright, I have now officially tracked down the source of the problem.

The FreeType 2.1.9-1 source package that comes with CentOS 4.2 is not built
properly, or at least packaged properly.

It includes a patch called "freetype-2.1.3-enable-ft2-bci.patch" that is
supposed to turn on the ByteCode Interpreter within FreeType.

This patch as you see dates back to version 2.1.3 ... And when applied to
the 2.1.9 source code, it doesn't work (the line numbers don't match?),
which means that the 2.1.9 binary package does NOT have the Bytecode
Interpreter enabled, as other releases always did.

In my case at least, this causes the rendering problems I've been
experiencing.  Also because I wasn't aware of this option, any package I
compiled manually would suffer from the problem as well.

If I recompile the 2.1.9 source and manually edit the #define for BCI, then
everything is back to normal!

Bloody hell.

I suppose not all font files get affected the same way by this, so not every
one might notice a difference?

Cheers,
J.F.

-----Original Message-----
From: Steven Bowden [mailto:steveb at bundaberg.qld.gov.au] 
Sent: March 5, 2006 5:57 PM
To: Doyon, Jean-Francois
Cc: MAPSERVER-USERS at lists.umn.edu
Subject: Re: [UMN_MAPSERVER-USERS] MapServer incompatibility with RHEL4/CE
NTOS4.2

No Problems with mapserver-4.6.2 on CentOs 4.2 here on a Dual Code AMD 64
running 32bit;

I am using
proj-4.4.9
gd-2.0.33
gdal-1.3.1
geos-2.2.1
mapserver-4.6.2
php-4.4.0

gcc 3.4.4
glibc-2.3.4-2.13
freetype-2.1.9-1

Mapserver config
./configure --with-postgis --without-tiff --with-jpeg --with-threads
--with-proj=/usr/local --with-gdal=/usr/local/bin/gdal-config
--with-php=../php-4.4.0 --with-gd=/usr/local --with-freetype=/usr/bin
--with-ogr=/usr/local/bin/gdal-config --with-wms --with-wfs --with-wcs
--with-geos=/usr/local/bin/geos-config --enable-debug

Regards
Steve

On Sat, 4 Mar 2006 2:58 am, Doyon, Jean-Francois wrote:
> I tried moving the RedHat 7.3 binary that I know works fine to a 
> CentOS 4.2 system, and wham, the problem appeared!
>
> I also investigated trying to run multiple glibc's to see if that 
> might be the problem (i.e. install the RedHat 7.3 glibc on CentOS for 
> the use of MapServer), but that looks like quite the nightmare to try out
...
>
> J.F.
>
> -----Original Message-----
> From: UMN MapServer Users List [mailto:MAPSERVER-USERS at LISTS.UMN.EDU] 
> On Behalf Of Doyon, Jean-Francois
> Sent: March 3, 2006 10:28 AM
> To: MAPSERVER-USERS at LISTS.UMN.EDU
> Subject: Re: [UMN_MAPSERVER-USERS] MapServer incompatibility with 
> RHEL4/CE
> NTOS4.2
>
> OK, we're running 32 bit CentOS 4.2, on Xeon processors.
>
> I'm thinking there's some problem with the newer glibc or something 
> like that, which would also explain the crashes I was having at first 
> on a specific map?
>
> J.F.
>
> -----Original Message-----
> From: UMN MapServer Users List [mailto:MAPSERVER-USERS at LISTS.UMN.EDU] 
> On Behalf Of Doyon, Jean-Francois
> Sent: March 3, 2006 9:56 AM
> To: MAPSERVER-USERS at LISTS.UMN.EDU
> Subject: Re: [UMN_MAPSERVER-USERS] MapServer incompatibility with 
> RHEL4/CE
> NTOS4.2
>
> Lowell,
>
> Indeed! These are 64 bit CPU's! (Intel Xeon, Dell PowerEdge 1850's)
>
> And we're running CentOS, though I'm not sure whether it's the i386 
> version, or the x86_64 one, I'll look into that.
>
> I've also replicated the problem on a Pentium 4 machine though, which 
> is 32 bit, though with what intel calls EM64T technology.
>
> Could it be the HyperThreading?
>
> J.F.
>
> -----Original Message-----
> From: Lowell.Filak [mailto:lfilak at medinaco.org]
> Sent: March 3, 2006 9:41 AM
> To: Doyon, Jean-Francois
> Cc: MAPSERVER-USERS at LISTS.UMN.EDU
> Subject: Re: [UMN_MAPSERVER-USERS] MapServer incompatibility with
> RHEL4/CENTOS4.2
>
> Doyon, Jean-Francois writes:
> > Hello,
> >
> > OK well I've tracked down the source of my problem to the operating 
> > system itself it seems.
> >
> > Regardless of the versions/combinations of mapserver/gd/freetype I 
> > use, there are definite visual rendering problems when on this OS.  
> > On RedHat 7.3, everything is fine!
> >
> > Example:
> >
> > http://atlas.gc.ca/family-good.gif
> > http://atlas.gc.ca/family-bad.gif
> >
> > Note the small dot that appears on the symbol?
> >
> > Also:
> >
> > http://atlas.gc.ca/the-market-good.gif
> > http://atlas.gc.ca/the-market-bad.gif
> >
> > Notice how on the bad version, a bunch of point symbols are missing?
> >
> > I tried various combinations of:
> >
> > MapServer 4.6.1 and 4.8.1
> > FreeType 2.0.9 and 2.1.9
> > GD 2.0.28 and 2.0.33
> >
> > A given combination works on 7.3, but not on CentOS.
> >
> > Without luck.  The problem with the strange shape of the symbols ( 
> > http://atlas.gc.ca/mapserver-symbol-problem.png ) seemes to be 
> > related to the version of FreeType for some reason?  FreeType 2.0.9 
> > acts as we expect it to, FreeType 2.1.9 does not.  There may be 
> > cumulative effects
>
> also.
>
> > CentOS 4.2 uses GCC 3.4.4 and GLIBC 2.3.4.  I had gotten a horrible 
> > crash message at some point with a version of mapserver on this OS
> > (http://lists.umn.edu/cgi-bin/wa?A2=ind0602&L=mapserver-users&P=R608
> > 02 &I=-3& X=106ECE1CAE6B73E23F), but upgrading from 4.6.1 to 4.8.1 
> > had fixed it.  This was related to memory allocation.  Could 
> > something related to that be causing this?  Turns out I've got 4.6.1 
> > running on this platform now, but that really didn't help.
> >
> > I'm out of ideas :( Anyone?
>
> J.F.,
>
> Is the 7.3 box a 32-bit and the RHEL box a 64-bit kernel by any chance 
> (you may have already stated that in your previous "double free 
> detected.." post but I don't remember). Not that I know of a 
> particular fix, but we have certainly had strange problems here also.
>
> Lowell