How are installation sets being built?

Hi all,

I was always thinking that --enable-epm (despite its name) was the magic 
autogen.input switch to instruct a top-level "make" (or "make check") to 
also build installation sets (deb, rpm, dmg, msi, ...), on all 
platforms.  (And that the phenomenon that Windows continued to build an 
archive installation set regardless of --dis-/enable-epm was an 
inadvertent leftover from the old OOo times where such an archive 
installation set was needed to run the subsequentcheck tests agaist.)

However, "git grep -Flw -e --enable-epm" makes me doubtful about my 
assumption, as only

distro-configs/LibreOfficeLinux.conf
distro-configs/LibreOfficeMacOSX.conf
distro-configs/OxygenOfficeLinux.conf

mention it, while

distro-configs/LibreOfficeAndroid.conf
distro-configs/LibreOfficeAndroidX86.conf
distro-configs/LibreOfficeMinGW.conf
distro-configs/LibreOfficeOpenBSD.conf
distro-configs/LibreOfficeWin32.conf
distro-configs/LibreOfficeWin64.conf
distro-configs/LibreOfficeiOS.conf
distro-configs/OxygenOfficeWin32.conf

do not.

So, I am wondering how exactly the official TDF-released installation 
sets are being built for the various platforms.  What are the autogen.sh 
(autogen.input) switches, and what are the exact make command lines? 
Where is that stored in git or similar (if it is stored anywhere at all)?

Similarly, how exactly are any of those "nightly" installation sets 
(uploaded from select tinderboxes, IIUC) being built?  How can you trace 
that?

(I'm asking because I noticed that after instdir-ification we still 
always build a now-useless archive installation set on Windows during 
"make" and "make check", regardless of --dis-/enable-epm, and I wanted 
to clean that up.  Which is especially important now that that 
installation set is built in parallel with running the "make check" 
subsequenttests, which appears to have destabilizing effects at least 
for some.)

Stephan