[PATCH 1 of 7 V2] [pyramid/ctl] pyramid command will always run in foreground from now on

Laurent Peuch cortex at worlddomination.be
Tue Sep 24 17:34:07 CEST 2019

On Tue, Sep 24, 2019 at 05:29:43PM +0200, Laurent Peuch wrote:
> On Tue, Sep 24, 2019 at 03:31:07PM +0200, Denis Laxalde wrote:
> > Laurent Peuch a écrit :
> > > Turns out it is actually fixed in this series in changeset
> > > c1d201581b16 "[cubicweb-ctl] backport --loglevel option of pyramid to
> > > all instance commands", the merge_options was missing in pyramid.
> > > 
> > > Dunno if you want a specific patch for that one?
> > > 
> > > The other error on ^C for pyramid (when pyramid command returned 1) is
> > > also solve in the pdb related changeset.
> > 
> > Yes, it's not that obvious to me which patch to pick to solve this issue
> > from the last series you sent (i.e. I'm lost).
> > 
> > So if you could send a series of patches to solve this issue first and
> > independently of any features, that'd be really nice.
> Well, it's just the merge_option of this changeset
> https://hg.logilab.org/users/lpeuch/cubicweb/rev/c1d201581b16#l2.26
> Which is in this email
> https://lists.cubicweb.org/pipermail/cubicweb-devel/2019-September/040893.html
> I can really send just this thing extracted of this other patch if you
> want, but rebasing/evolving/spliting again and again and again
> changeset that I've been working months ago really end up introducing
> bugs in the process and as an illustration of that: this bug is
> actually a result of that.
> > On the other hand, I'm not sure the proposed solution (merge_options) is
> > the way to go. Some times ago, I questioned the usefulness of the --pdb
> > flag for /all/ instance commands. Specifically, what's the point of
> > having this for the "pyramid" command?
> You never when, where and how you are going to have a bug: maybe
> pyramid server is going to crash, maybe it's going to crash before for
> other reasons like confusing bug in the configuration, or you end up
> in a weird conner case that doesn't seems to make sens, you just never
> know upfront and adding this debugging feature is cheap and easy and
> you never know when it will save you time, you just want it everywhere
> ready even when it's not used often, that still makes things easier
> for everyone.

Nevertheless, if just having a warning is that annoying for you, we
can try to find a compromise like adding a new "--debugtoolbar" option
or another solution.


Laurent Peuch -- Bram

More information about the cubicweb-devel mailing list