Changes between Initial Version and Version 1 of Ticket #51516, comment 153


Ignore:
Timestamp:
Jun 11, 2024, 11:12:38 PM (3 weeks ago)
Author:
RJVB (René Bertin)
Comment:

Legend:

Unmodified
Added
Removed
Modified

  • Ticket #51516, comment 153

    initial v1  
    1 I now have a first draught of a minimally patched "base" source tree that *almost* builds. I've had to give up the idea to build only the strictly necessary components as it isn't clear exactly how to make those available, IMHO that's definitely something that could be worked on in a follow-up effort. Esp. if we can work with the assumption that any headers required should already be installed, for instance.
     1And here we are:
    22
    3 I'm currently running into what appears to be an incompatibility in readline.c, where at least one of the completion function `#define`s is misdefined. Is there something I should know here re: using port:libedit vs port:readline (I notice that my original Pextlib links to the system libedit on Mac, but libreadline on Linux...)
     3https://github.com/RJVB/macstrop/tree/master/sysutils/MacPorts-pextlib
     4
     5For now this has only an implementation for the `MacPorts-pextlib-devel` subport which follows the MacPorts master branch and has my personal patches for Linux and features - in addition to (and distinguishable from) a number of patchfiles that are required to get things to build against installed ports. I've also figured out how to build only in the few directories that are of direct interest without having to build the entire project. This was as simple as using `port:sqlite3` as a build dependency rather than building the copy hidden in the vendored Tcl source tree. There are no hidden costs to it either; port:sqlite3 doesn't add new, indirect dependencies.
     6
     7I tested (on Mac and Linux):
     8- port fetch
     9- port checksum
     10- a rebuild and reinstall (`upgrade --force`) of `MacPorts-pextlib-devel` itself
     11
     12One of my feature patches reintroduces the Tcl binding for `mkstemp` that was dropped, so I could test with an older "base" install. Being able to do this is not a goal as discussed above but I think that the fact that everything I tested works normally indicates that the approach used here is reliable.