[Lazarus] Inno Setup no longer supports Windows 95, Windows 98, Windows Me

[Marco van de Voort]

On Fri, Aug 03, 2012 at 06:18:24PM +0200, Sven Barth wrote:
> > The main difference though is that they concern different targets, and do
> > not share a Tier 1 target like win9x
> 
> Nevertheless we have (official) testers for those two targets. They take 
> care for them and even add new features (remember Tomas adding support 
> for TProcess to OS/2 some months ago?). That's currently more love than 
> Win9x gets...
> 
> Also FreeDOS (on which FPC can run) is an actively developed/distributed 
> platform like is ECOMStation. This is again unlike Win9x.
> 
> So in my opinion DOS and OS/2 should not be put into the same bin as Win9x.

The category was defined as being mostly active during release engineering.

Dos and OS/2 (but also e.g. Debian) are together responsible for 80% of the
post branching and pos RC1 activity. That is disproportionately much.

Win9x has fewer such issues in number, probably because it piggy backs on a
major target, but often its issues slip past RC1, causing a Tier 1 target to
be rebuild without RC validation (for IMHO next to nothing). At least OS/2
and Dos only risk their own releases.

This is the major reason for the splitting proposal. (

Though again, IMHO we should simply stop support win9x in main builds. 
Interested people can maintain 3rd party builds for a while for the unhappy
few.  Technical possibilities (slapping external unicode libraries under it
etc etc) enough, but IMHO that will all be talk and tinkering, but will
never reach a realistic level of releasable quality.

If people show they can maintain a 3rd party port to win9x for a whole major
cycle, we can always revert and split at that point. But IMHO first show is
viable, then infrastructural changes.