Important information regarding latest Xorg update packages

[Tony Nelson]

At 10:30 AM -0400 9/23/05, Gene Heskett wrote:
>On Thursday 22 September 2005 23:16, Chris Adams wrote:
 ...
>>"#" seems to work as a comment because many of the sections of spec
>>files are run through a shell or other scripting language that
>>recognizes "#" as a comment; rpm passes the lines through untouched.
>>If you have a line like:
>>
>># %define xyzzy 123
>>
>>then the define takes place; rpm doesn't see "#" as a comment.
>>
>>The problem lines from the spec file in question:
>>
>>************************************************************************
>>%postun Mesa-libGLU     -p /sbin/ldconfig
>>
>>##### xfs scripts ####################################################
>># Work around a bug in the XFree86-xfs postun script, which results in
>> the # special xfs user account being inadvertently removed, causing xfs
>> to run as # the root user, and also resulting in xfs not being
>> activated by chkconfig, # This trigger executes right after the
>> XFree86-xfs postun script, and ensures # that the xfs user exists, and
>> that the xfs initscript is properly chkconfig # activated
>> (#118145,118818)
>>%triggerpostun xfs -- XFree86-xfs
>>************************************************************************
>>
>>The spec file syntax says that the "-p <program>" arg to %postun means
>>to run the script with <program> instead of /bin/sh.  If you don't
>>provide a script, the program is run with no input.  This is great for
>>something like (obviously useless example, but you get the idea):
>>
>>************************************************************************
>>%postun -p /usr/bin/perl
>>$x = 1;
>>print $x + 1, "\n";
>>************************************************************************
>>
>>However, what happens in the xorg problem spec file above is that rpm
>>takes everything up to the next spec file directive as the script (only
>>if the following lines are empty does rpm treat it as "no script").
>>Since "#" is not a spec file comment, that means rpm thinks that the
>>"##### xfs scripts" lines before the %triggerpostun is actually the
>>%postun script for xorg-x11-Mesa-libGLU, and that it should be run
>>through /sbin/ldconfig.  Of course, ldconfig doesn't know what to do
>>with it; that's where we get the error.
>
>Oh cute...
>
>Where is there a bible for writing spec files?  Or, better yet, why
>isn't rpm re-written so a # in char position 1 is always a comment
>line?  That seems like the sensible thing to do, but I'm but the
>expert...

It might be safer (and simpler to get done) to enhance rpmlint to warn of
such things:  places where just having an apparent comment changes what
will happen (non-empty input will be sent to the comand instead of empty
input); and "scripts" that don't use a recognized script interpreter, with
some popular hacks specifically recognized as unsafe.  I suppose I could
add a RFE for rpmlint.
____________________________________________________________________
TonyN.:'                       <mailto:tonynelson at georgeanelson.com>
      '                              <http://www.georgeanelson.com/>