Re: [ptxdist] [PATCH 3/7] ptxd_make_world_get: make 'ptxdist urlcheck' idempotent

[Roland Hieber <rhi@pengutronix.de>]

On Fri, Jun 19, 2020 at 01:07:22PM +0200, Michael Olbrich wrote:
> On Fri, Jun 19, 2020 at 10:22:55AM +0200, Roland Hieber wrote:
> > On Fri, Jun 19, 2020 at 07:34:01AM +0200, Michael Olbrich wrote:
> > > On Sat, Jun 13, 2020 at 03:32:45AM +0200, Roland Hieber wrote:
> > > > There is no guarantee that URLs will still work the next time we want to
> > > > check them, so try to fetch the URL every time instead of caching the
> > > > result. This also makes iterations of editing the URL in the rule file
> > > > and testing if it works quicker.
> > > > 
> > > > Unfortunately, GNU make skips implicit rules for .PHONY targets, but it
> > > > is unlikely that $(STATEDIR)/%.urlcheck is created by something other
> > > > than PTXdist, so removing the $(call touch) is a good approximation for
> > > > phony behaviour.
> > > 
> > > No. This is deliberately like this. I want to run 'ptxdist urlcheck', fix
> > > errors and then continue with the rest.
> > 
> > Can you not use -k to do this?
> 
> I do that. But without the successful cache, I need to track all failures
> manually because I cannot call 'urlcheck' again. To check the rest.

Ah, I now understand what you mean. Of course, successfully checked URLs
should be cached, but my change breaks that. My intention with this
patch was that the failed urlchecks are not cached if they are
unsuccessful. But somehow I can no longer reproduce that problem, an
unsuccessful urlcheck now fails without creating the state file... so
I guess everything should be fine without this patch.

> > Caching an error state is at least unusual, and it is not done anywhere
> > else in PTXdist, so this breaks the principle of least surprise.
> 
> On the contrary. In general, if something was successful, then we don't do
> it again unless there is a clear indication that the input has changed.
> In all other cases an explicit 'drop' or 'clean' is needed.
> 
> For example, we don't skip the 'touch' for local_src/ links either. Even
> though the code can change any time.

I meant, if a package fails to compile or install, we don't cache that
state, so the next ptxdist go will try to build it again. Of course, not
all error states result in a compilation failure, but we can do the best
and fail for those that do.

 - Roland

-- 
Roland Hieber, Pengutronix e.K.          | r.hieber@pengutronix.de     |
Steuerwalder Str. 21                     | https://www.pengutronix.de/ |
31137 Hildesheim, Germany                | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686         | Fax:   +49-5121-206917-5555 |

_______________________________________________
ptxdist mailing list
ptxdist@pengutronix.de
To unsubscribe, send a mail with subject "unsubscribe" to ptxdist-request@pengutronix.de