Le mardi, 25 décembre 2012 21.45:41, Julien Cristau a écrit : > On Fri, Dec 7, 2012 at 22:31:33 +0100, Didier Raboud wrote: > > Please unblock package cups (…) > > Questions: Hi Julien, thanks for this review! > - what does "mv /etc/cups/cupsd.conf /etc/cups/cupsd.conf.conffile-bak" > in preinst achieve? In versions prior to 1.5.3-2.7, /etc/cups/cupsd.conf is a conffile, registered as such by dpkg. So in a normal upgrading case, the file would either be dropped (if never modified) or marked obsolete (if ever modified). As we want neither, moving it away before the dpkg unpack (making this file unnoticed to dpkg) and reinstating it after the unpack (in postinst for normal cases and postrm "abort-upgrade" in case something fails) makes /etc/cups/cupsd.conf a non-dpkg-conffile without having it marked as "obsolete" (which is irrelevant for that file). Granted, /etc/cups/cupsd.conf should not be in /etc/ but in /var/lib/, but that would be the subject for another intrusive patch (and upstream apparently plans to do that change in a not-too-distant future). > - in cups-dbus-utf8.patch, I'm wondering if the "if (str_len > buflen)" > check isn't off-by-one? To be honest, I don't really know. As I read the code (and the surrounding comments), making it off-by-one puts the buffer size on the safe side (aka always at least as large as the input string). > Actually, one more: the validate_utf8 thing uses en_US.UTF-8, which may > or may not exist. You might want to use C.UTF-8 instead, which is in > our libc since 2.13. Good catch! I'll upload a new cups 1.5.3-2.11 with that fix. What is the good way to Depend on libc > 2.13 across all our architectures ? Cheers, OdyX
Attachment:
signature.asc
Description: This is a digitally signed message part.