[fpc-devel] [RFC] Modernising the FPC Release Process -- Proposal for Review

[Marco van de Voort]

Op 15-4-2026 om 17:13 schreef Tomas Hajny via fpc-devel:
>
> First of all - I understand that you try to move things forward and I 
> appreciate your interest. In addition, I understand that you prefer 
> start working on something than participating in potentially endless 
> discussions. However, I believe that you should read at least some 
> parts of that discussion. 

Current distance between fixes and main is about 13739 revisions. (not 
counting the 23 fpcbuild revisions, and the work that needs to be done 
for docs). Of course a MR can be more commits/revisions, but rarely more 
than 20. So 100 merge requests is at most 2000, and that is already a 
stretch, both wrt commits/mMR and  that that already assumes that are 
most are viable to merge and will be eventually merged.

As I said in the previous mail, the MR surplus is a new problem that 
still is developing. It is IMHO not saying much about the project, 
except that we as a project, we have the tendency to let revisions or 
bugreports linger rather than close them without resolution.

Also I'd expect many far-out MR only to be merged after the branching of 
a new stable branch, as that would provide maximum time to test and 
stablish them in main/trunk, so a high number of unmerged MRs so late in 
the cycle seems normal. They are just more easily counted, rather than 
being private branches of developers or patchsets in bugreports as in 
SVN times.

Anyway some 500-1000 of those 13700 revisions are makefile regens and 
verrsion updates that can be somewhat ignored. Also anything attached to 
fcl-web or fcl-passrc/pas2js has essentially not been merged for 3-4 
years or longer due to high mutual entanglements and high development 
speed due to pas2js.

Note that while the current numbers and wall time is extreme, an higher 
number of  unmerged non compiler revisions is normal for being in the  
x.y.4 release cycle.

Compiler revisions are a different breed altogether, and only very 
sporadically merged. So much that the merged revisions mostly disappear 
in the noise (I'd guess 100-200 revs, and most of them relatively soon 
after 3.2.0 out of  a total number of compiler related revisions of 6000+).

For a simple analysis by category see 
https://www.stack.nl/~marcov/mergelogs32/   updated every few weeks, and 
https://www.stack.nl/~marcov/mergelogsbuild32/ for fpcbuild/