[OpenIndiana-discuss] login issue since upgrade to illumos-5c4a5fe167

[Alan Coopersmith]

On 10/23/24 15:22, Marcel Telka wrote:
> On Thu, Oct 24, 2024 at 12:13:11AM +0200, Marcel Telka wrote:
>> On Wed, Oct 23, 2024 at 03:09:23PM -0700, Alan Coopersmith via openindiana-discuss wrote:
>>> On 10/23/24 14:03, Peter Tribble wrote:
>>>> One other thing that occurs to me: does OpenIndiana still run a find to see
>>>> if the
>>>> cache needs to be regenerated due to new loaders? I largely removed all
>>>> that logic
>>>> because (a) it was quicker to refresh the caches than work out if it was
>>>> necessary,
>>>> and (b) it was susceptible to timestamps getting out of whack.
>>> We did too:
>>> https://github.com/oracle/solaris-userland/commit/87c2d590ff309b21a923d75dd5533ffacc04777f
>>
>> Nice.  I really love it! :-)
> 
> And you even torched the desktop-cache component three years ago.

Oh yes, having that as a separate package for just the SMF manifests and method
scripts that didn't include the commands they ran, bugged me for a long time,
until I finally got around to fixing it (which required first nuking the
"find_newer" script and switching to "find -newer" so I didn't have
to make all the other packages depend on a tiny find_newer package):

https://github.com/oracle/solaris-userland/commit/ba967091f5e75913a3a273291882008516c0a239
https://github.com/oracle/solaris-userland/commit/9f1a882f986b35fb81575a6ec3ec674697739cd0

The bug report for the first notes:

"The find_newer script seems to be a workaround for SVR4 packaging preserving
the timestamps on installed files, thus defeating the check if files had
been updated, since they'd often have an older mtime, but newer ctime.
  
This should not be needed any longer with IPS, since we don't preserve
timestamps on the files that are inputs to the caches so can rely on the
mtime to find updates.

/usr/share/desktop-cache/restart_fmri was a script for SVR4 packages to
call to do the equivalent of the IPS restart_fmri actuator."

	-alan-