gcc43 installed corrupted receipt; made MacPorts unusable

Ryan Schmidt ryandesign at macports.org
Thu Sep 13 03:43:09 PDT 2007

So we had a problem with the gcc43 port (until it was fixed in  
r29009) where just attempting to install gcc43 rendered the entire  
port system unusable, because MacPorts couldn't process the receipt  
that got installed. See this ticket:


The problem occurred when variant descriptions were added to the port  
like this (in r28362):

variant odcctools
         description "some description" {

Those of us less familiar with tcl syntax rules (me; probably most of  
us) probably didn't catch that this should have been written:

variant odcctools \
         description "some description" {


variant odcctools description "some description" {

I don't know enough about tcl to understand why the original version  
was unacceptable, and in particular, if it was unacceptable, why it  
didn't blow up much earlier. What surprised me was that "port lint"  
didn't find anything wrong, and that "port install" didn't complain  
at all about the port, until the entire thing exploded during the  
install phase. "port info" gave a slight clue when it printed  
"(Variants: universal, darwin_8, powerpc, )" (the two variants didn't  
show up, but there was a comma and a blank after the powerpc  
variant). "port variants" similarly didn't show the two variants, but  
had an extra blank line at the end.

What can we do to prevent portfiles malformed in this manner from  
breaking MacPorts in the future? It seems that there are parts of the  
code that are assuming things that are not true in portfiles  
malformed in this way. Perhaps to be more robust we should add more  
sanity-checking code to base. But I'm not sure where or how.

More information about the macports-dev mailing list