Re: [PATCH 3/3] SConstruct: Use 'target' option to decide whenewer to restart systemd

[Ladislav Michl]

Hello Gary,

On Tue, Feb 04, 2020 at 02:16:43PM -0800, Gary E. Miller wrote:
> Yo Ladislav!
> 
> On Tue, 4 Feb 2020 01:37:06 +0100
> Ladislav Michl <address@hidden> wrote:
> 
> > > I have updated "scons -h" to try to make the "target" and "sysroot"
> > > options more clear.  
> > 
> > Typo "toolchaons".
> 
> Thanks.  Fix pushed.
> 
> > And it is really directory? Seems to be prefix:
> 
> Can you cite a toolchain that uses it as a prefix?  Is so, please suggest
> improved help text.

Why did you delete the most relevant part of my answer when replying?
I can point you to the Pathname Resolution as defined by POSIX:
https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap04.html#tag_04_13

SConstruct takes target, adds hyphen and concatenates with toolname.
How does that match with definition of a directory?

> > Yes, that is a valid point. I just didn't want touch too much, so I
> > used scons option "target". It is already used the same way when
> > deciding time_t size (SConstruct reads):
> >     if env['target']:
> >         announce("Not checking sizeof(time_t) when cross-compiling")
> >         sizeof_time_t = 8
> 
> Once again, not necessarily cross-compiling.  Just as likely to be
> a secondary run environment.

Same as providing "sysroot" does not necessarily mean cross-compiling,
but the patch at least makes it consistent.

> > > Maybe gpsd needs an "I am cross-compiling" option?  
> > 
> > I'd rather avoid adding new options, since daemon reaload is meant
> > to ease development here, so (mis)using 'target' seem less evil
> > thing to do.
> 
> I also avoid new options, but it is better than mis using existing ones.
> 
> Why do you care the "systemd reload" gets run?  Zero side effects.

What if anyone else is in the middle of editing systemd files and daemon
reloads invalid configuration? Eh, why do I arguing, doing so is just
insane.

> > Searching for cross compilation using scons I came to
> > https://github.com/SCons/scons/wiki/PlatformToolConfig
> > Is it still only a draft?
> 
> I have no idea.  In this case, we are talking about the gpsd SConstruct,
> so that does not apply.

Unfortunately. It would take ages until gpsd SConstruct reaches at least
state of this draft.

> > > As it stands now, if systemd if not found, or the scons options
> > > systemd=False is set, then the "systemctl daemon-reload" is not run.
> > > 
> > > I assume that you want systemd=True, so that the systemd files get
> > > installed in DESTDIR.  But you do not want "systemctl
> > > daemon-reload" to run.  
> > 
> > Yes, that is my use case.
> 
> Why do you care?  Zero side effects.

Not true.

> > > Reloading the daemon is benign, having zero bad side effects.  But
> > > for OCD purposes, maybe improving the situtation is worth effort.  
> > 
> > It does not happen at all, running scons as ordinary user. I guess
> > it should also not happen when installing to DESTDIR, because then
> > nothing changes for systemd.
> 
> Once again, what's the problem?

systemctl daemon-reload || true
Failed to reload daemon: Access denied

But before that dialog asking for root password pops up.

I'm well aware 'udev-install' rule is originally for testing, if you think
it should remain usefull only for testing, I'm fine with that.

Best regards,
        ladis