On 2026-04-16 10:38, Michael Van Canneyt via fpc-devel wrote:
On Thu, 16 Apr 2026, Tomas Hajny via fpc-devel wrote:
.
.
If I understand it correctly, this part of the proposal suggests
replacing the current mechanism of keeping fpcbuild repository up to
date (i.e. keeping the latest versions of the submodules always
available in the fpcbuild repository) with one time synchronization
only when a particular tag is applied to fpcbuild. That would mean
that building the most up to date version for individual targets could
not be tested before applying the tag (and that testing it again after
any applied fixes would always need manual synchronization). And that
would not be a good idea at all as far as I can tell...
1. How is this different from today ?
You will always need a manual action to 'fix' the version on the 3
repos involved.
Whether that is setting a tag or changing the module branch on the
.submodules file,
it remains an action that needs to be done.
Today's action for getting all needed sources is "git pull
--recurse-submodules" in a git clone/checkout for the particular branch.
2. What stops people from testing using fpcbuild and whatnot without a
tag being set ?
Nothing. The branch from which the release is being built is known
long
before any tag is set, so nothing stops people from testing that
branch.
To test, you need 3 SHAs: fpcbuild, source and docs. That's it.
We put this in a build_id_card.txt or whatever which you can modify
and you're all set
to launch any build you desire.
My only suggestion was to prevent a step which seemed to make testing
the builds more complicated. You miss my point that "3 SHAs" are a
moving target during the release building testing.
Honestly, I don't understand all the negativity. If there is a problem,
we'll fix it.
All the negativity? I only tried to point out specific weak points of
the proposal. I left major parts without any comments (except a general
statement that I appreciate Graeme's interest in moving things forward,
which shouldn't be considered negative hopefully ;-) ). I tried to
contribute based on my specific experience. I believe that Graeme
understood it that way, BTW.
Tomas
_______________________________________________
fpc-devel maillist - [email protected]
https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel
