On 16May26:1221+0200, Willy Goiffon wrote:
Erich Eckner wrote:
On 19.10.2015 13:58, Willy Goiffon wrote:
Bernd Eggink wrote:
On 18.10.2015 17:20, Willy Goiffon wrote:
I had an idea to make software patching easier, and lower the need to fork ports.
But you still would have to maintain an additional source (your patches) - whenever the original port changes, a patch may become obsolete, not applicable, or plainly wrong. I never include ports from contrib directly, but copy them into my own repository, update the version (most of them are outdated), adjust the Pkgfile to my needs, make patches etc. This has the advantage that everything that belongs to the port is concentrated in a single directory.
Sure you'd have to maintain the patches yourself, but you'd have to do it too in case of a fork.
sry to bring up the past, but I recently realized I have a few ports myself which are only cloned due to some simple patches, e.g.: - additional directive to ./configure in curl - very short patch for conky to display upto 20 instead of 10 processes - ...
Thus I modified /usr/bin/ports itself to do something similar as proposed by Willy.
This looks like a 'higher-level' approach to me. My idea was to provide a way to patch the software directly, while yours is provides a way to patch the whole port (Pkgfile/md5sum).
I follow Mother's rule number 1 in "What Mother Never Told You About VM Service" (http://www.leeandmelindavarian.com/Melinda/tutorial.pdf), that still holds a lot of good basic sysadmin rules even though most of the specifics are different: "Never change anything IBM sends you." In context that means keeping IBM's file contents and organization unmodified and making copies elsewhere and editing the copies, then ensuring that maintenance processing uses your copies instead of IBM's. The VM/CMS disk search order facility makes that easy and I think inspired Plan 9's union directory facility. I apply this to CRUX by staying the heck out of the canonical repositories (I'm not certain what rsync is going to do to any changes I make in there, anyway). This also means I can just build anything exactly as the distribution intended simply by temporarily inserting an override /etc/prt-get.conf prtdir for the original directory:package as the first prtdir; e.g., prtdir /usr/ports/opt:grub2 The point is I know it's a canonical build if the port resides in an official port collection, I never have to answer the question, "Was this build tainted in some way?" (other than my environment's idiosyncracies, of course). If I need to modify a distributed port (who doesn't ever need to do this?), I cp -a copy that port into a non-distributed collection (I use /usr/ports/local) and that is normally at the head of the prtdir records. I then edit the Pkgfile and, as Mother said in rule number 4: Always leave tracks, I add a "Localized:" header documenting the raison-d'etre; e.g., tmpfix, the source collection, the target system(s), and the localizer (me). Below the headers I add commentary as needed to explain why this localization was made. A diff between the two Pkgfiles makes it very clear what is afoot. In regard to local patches, they are no different than official patches except they exist in a non-distributed collection. If a patch applies to multiple ports, you can symlink to one actual patch but need to put them into a fixed location (I just copy such patches as needed since I move retired local ports from local to local.inactive which appears not in /etc/prt-get.conf). -- <not cent from sell> May the LORD God bless you exceedingly abundantly! Dave_Craig______________________________________________ "So the universe is not quite as you thought it was. You'd better rearrange your beliefs, then. Because you certainly can't rearrange the universe." __--from_Nightfall_by_Asimov/Silverberg_________________