[pca] Patches with passwords

Martin Paul martin at par.univie.ac.at
Thu Mar 27 12:21:33 CET 2008


Laurent Blume wrote:
> The same person posted a list of patches that he said hanged in the
> postpatch part, because they were asking for a password:
> 
> I can't test them myself, since I don't use those products.

Me neither, but I had a report about that behaviour with 119325 last 
year. I did some research back then, and found that there's no simple 
way to deal with these patches. They do not have the "interactive" 
property set in their patchinfo files, which could be used to identify 
these patches.

As pca buffers (and usually doesn't display) the output of patchadd, 
there's no way that that user running pca will see a prompt issued by 
the postpatch script.

So my workaround in v5.7 was to modify pca to run patchadd with 
"</dev/null", assuming that the postpatch script will fail as it does 
get no/wrong input. In that case pca will show output from patchadd, so 
the user would at least see that the patch requested input, and could 
install the patch manually.

 > Do you deal with that kind of peculiarity?

In my opinion, there's no easy workaround, and the patches should be 
fixed by Sun. There *are* patches which require post-installation things 
to be done, but for most patches these are explained in the README. 
There are postpatch scripts for things that can be done automatically, 
but besides those in the above mentioned patches, they run hands-off.

If Sun really decides that patches which require user-input during 
installation are allowed, they should at least be marked as 
"interactive" correctly, for patch tools being able to behave properly.

Martin.



More information about the pca mailing list