Source-Changes-D archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: CVS commit: src/usr.sbin/sysinst/arch



On Sat, Jun 18, 2022 at 12:22:44AM +0200, Joerg Sonnenberger wrote:
> > - ramdiskbin with CHECK_ENTROPY requires extra 7000 bytes, at least on
> >   m68k, so poor tier-II ports that have size restrictions on install media
> >   or use slow floppies rather prefer smaller binaries
> 
> Hm. It shouldn't be anywhere as heavy. Can you check where that size
> comes from? 1KB I can understand, but 7KB feels too large and something
> is going wrong. E.g. it could be pulling in more SHA2 code that hasn't
> been used before and there might be cheaper options in terms of code
> size.

Many of that is from additional messages and menus, not code size:

Default build for atari:

   text	   data	    bss	    dec	    hex	filename
 315441	  20296	  30948	 366685	  5985d	sysinst
-rw-r--r--  1 root  wheel  37109 Jun 18 11:55 sysinstmsgs.de

CHECK_ENTROPY=no:

   text	   data	    bss	    dec	    hex	filename
 305976	  18084	  29924	 353984	  566c0	sysinst
-rw-r--r--  1 root  wheel  34651 Jun 18 11:58 sysinstmsgs.de


... which seems to mostly cover the crunched difference already, so likely
no (or not many) library/dependends effects in there.

Similar to what abs@ suggested, I have a long term plan to have multiple
(optional) sysinst phases, and we could move all late/configure stuff to a
separate phase and maybe actually reuse the new installed system binaries
for that. Other phases that will be needed are "update kernel and modules"
vs. "update everything else after reboot with new kernel". But this is
not going to happen for 10.

Martin


Home | Main Index | Thread Index | Old Index