[sane-devel] The bug of Canon-backends of SANE with Canocsan300by ISA SCSI

[Till Kamppeter]

I have made available debug-enabled Mandrake packages of SANE now:

sane-1.0.8-6mdk.src.rpm
libsane1-1.0.8-6mdk.i586.rpm
libsane1-devel-1.0.8-6mdk.i586.rpm
sane-backends-1.0.8-6mdk.i586.rpm
sane-frontends-1.0.8-2mdk.src.rpm
sane-frontends-1.0.8-2mdk.i586.rpm

Connor, and everyone else having a Canoscan connected with an ISA SCSI 
controller, please test doing the following steps:

Please download them from http://www.linuxprinting.org/till/tmp/ 
(immediately) or from a Cooker mirror (probably tomorrow) and put them 
all in one directory. Then upgrade with

    rpm -Fvh *sane*1.0.8-[26]mdk.i586.rpm

After having done so, configure your scanner and do

    scanimage -T

This will probably produce a file named "core" or "core.XXX" with "XXX" 
being an integer number (it must be the last entry of "ls -ltr core*"). 
Run the command

    gdb -c core.XXX

where you replace "core.XXX" by the name of your core file. Now you will 
have a "(gdb)" prompt. Type "backtrace" and press Enter. Then type 
"quit" and press Enter.

If you don have "gdb" on your system, install it with "urpmi gdb".

Send us all the screen output of both "scanimage -T" and "gdb". Try also 
other things and send us screen output of the program called and of "gdb".

If you have another RPMish distro than Mandrake, try to rebuild the RPMs 
for your distro:

    rpm --rebuild sane-1.0.8-6mdk.src.rpm
    rpm --rebuild sane-frontends-1.0.8-2mdk.src.rpm

Perhaps you must install a package named "rpm-build" or "rpm-rebuild" 
for that. You also need C compiler, Perl, and "make" on your machine. If 
more packages are needed you will get appropriate screen messages. Then 
proceed as shown above.

Please use "Reply All" in your mail program when replying to this, so 
that all people who got this mail also will get your answer.

Thank you in advance for your cooperation.

    Till


abel deuring wrote:
> Till Kamppeter wrote:
> 
>>That's no problem:
>>
>>Uninstall the SANE backends with
>>
>>   rpm -e --nodeps sane-backends
>>
>>Install "checkinstall":
>>
>>   urpmi checkinstall
>>
>>If this does not work, add a mirror carrying the contrib packages to
>>your rpmdrake/urpmi sources and try again.
>>
>>Get sane-backends-1.0.9pre1 or better the current CVS when the CVS
>>server is back again (Use this version because I plan to package SANE
>>1.0.9 for Mandrake 9.0, especially because I made the Epson Perfection
>>1260 running together with the maintainer of the "plustek" backend). Add
>>your code which you havn't checked in yet and use "checkinstall" instead
>>of "make install" to make it easier to de-install this SANE version and
>>cleanly install anotherone.
>>
>>Do not run "scannerdrake" any more now, and if you have a multi-function
>>device from HP also do not run "printerdrake". These programs try to
>>install the old packages again (and will choke due to the conflicts with
>>the SANE RPM which "checkinstall" has installed).
>>
>>The xsane provided by Mandrake Linux works also with SANE 1.0.9. If the
>>"scanimage" and "xscanimage" do not work, switch to
>>sane-frontends-1.0.9pre1 the same way as with the sane-backends-1.0.9pre1.
> 
> 
> Till,
> 
> am I right that you are Mandrake's maintainer for Sane? 
> 
> If you are, here are a few requests/comments:
> 
> According to the backtrace information posted by Henning on the mailing
> list, the segfault occurs somewhere in the sanei_scsi library.
> Unfortunately, the debug symbols seem to be removed from the binaries in
> the Mandrake rpm files, so we don't get very precise information about
> the location of the segfault. WHat Mitsuru and I would like to know is
> where exactly the segfault happens. So it would be best, if we and the
> users who have the Canoscan 300 and the other hardware, for which the
> segfault occurs, could get the "regular" Sane files from Mandrake 8.0 --
> but _with_ debug symbols, and also the source files, if Mandrake applied
> some patches to them. Using the most recent sources from Sane's CVS
> server could result in different behaviour.
> 
> Abel
> 
>