Please don't change the severity and priority settings on bugs, these are for developers to set in order to prioritise their time better.

These warnings should be fixed in Git master now. You could have just switched off fatal warnings though...

The gobject-introspection fatal warnings are not disabled using --disable-Werror -- they are set from a separate variable in the introspection.am makefile called SCANNER_WARN_ERROR.

This is set as so:

if OFFICIAL_RELEASE
SCANNER_WARN_ERROR =
else
SCANNER_WARN_ERROR = --warn-error
endif

which means that the only way to get this behaviour turned off is for this to be an "official release", which jhbuilding certainly is not.

Created attachment 59701 [details] [review]
remove leftover newline issue

The issues in telepathy-enums-gtk-doc.h are not fully handled by previous commit. Attached is a patch for the one remaining issue.

(In reply to comment #0)
> log of jhbuild break.

This...

(In reply to comment #4)
> remove leftover newline issue

... and this should already be fixed in git (0.18 and master).

(In reply to comment #3)
> The gobject-introspection fatal warnings are not disabled using
> --disable-Werror -- they are set from a separate variable in the
> introspection.am makefile called SCANNER_WARN_ERROR.

Yes, this is a bug. There should be a way to force them to be non-fatal, and the only use of $(OFFICIAL_RELEASE) should be its original purpose (deciding whether we require the lists of exported symbols to be complete, as they should be in every ABI-stable release, but not necessarily in between).

The default is to have various fatal warnings in every version that isn't a release tag - they get switched off during the release and switched back on afterwards - but there should be an override available for those, in both directions.

My suggestion is:

* for 0.18, where we won't add new API anyway: just make any future
  g-ir-scanner warnings never be fatal

* for master: add a --disable-fatal-warnings which switches off all the
  extras, and change jhbuild etc. to use that instead of --disable-Werror.
  This would include:

  * documentation completeness checks
  * coding style checks
  * default for --enable-Werror/--disable-Werror
  * default for g-ir-scanner --warn-error or not

Created attachment 59864 [details] [review]
[0.18] Don't make g-ir-scanner warnings fatal on this stable branch

Created attachment 59865 [details] [review]
[master 1/2] Merge the various "fatal warnings or not?" options

I've kept --disable-Werror (it's from a reusable macro for -Werror),
but distributors should now use --disable-fatal-warnings to control  
all the potentially-fatal checks simultaneously.

---

jhbuild counts as a distributor here, but so does, e.g., Debian or Fedora (if they plan to ship non-release versions, which I do not recommend - if you need a release, ask for one).

Created attachment 59866 [details] [review]
[master 2/2] Use --enable-fatal-warnings to select g-ir-scanner --warn-all

(In reply to comment #3)
> the only way to get this behaviour turned off is for this to
> be an "official release", which jhbuilding certainly is not.

For the record, this isn't strictly true - you could do:

    make SCANNER_WARN_ERROR=""

because of the strange way precedence works in make. I agree that that's not obvious or sufficient, though.

Comment on attachment 59864 [details] [review]
[0.18] Don't make g-ir-scanner warnings fatal on this stable branch

Review of attachment 59864 [details] [review]:
-----------------------------------------------------------------

Ok.

Comment on attachment 59865 [details] [review]
[master 1/2] Merge the various "fatal warnings or not?" options

Review of attachment 59865 [details] [review]:
-----------------------------------------------------------------

OK.

Comment on attachment 59866 [details] [review]
[master 2/2] Use --enable-fatal-warnings to select g-ir-scanner --warn-all

Review of attachment 59866 [details] [review]:
-----------------------------------------------------------------

OK.

(In reply to comment #5)
> (In reply to comment #0)
> > log of jhbuild break.
> 
> This...
> 
> (In reply to comment #4)
> > remove leftover newline issue
> 
> ... and this should already be fixed in git (0.18 and master).
> 
no this is for the one missed by previous fix . As reported in bug 48620 there is one newline left. and this patch fixes this. I had read the log and commit 5c58330c32f39a9dd6aaaef37f54461922cf6124 does not handles this one.

*** Bug 48620 has been marked as a duplicate of this bug. ***

(In reply to comment #13)
> > ... and this should already be fixed in git (0.18 and master).
> > 
> no this is for the one missed by previous fix

Thanks, also applied. Will be in 0.18.1 and 0.19.0.

(In reply to comment #10)
> [0.18] Don't make g-ir-scanner warnings fatal on this stable branch

Applied too. I reverted it for master.

(In reply to comment #11)
> [master 1/2] Merge the various "fatal warnings or not?" options
(In reply to comment #12)
> [master 2/2] Use --enable-fatal-warnings to select g-ir-scanner --warn-all

Both in master for 0.19.0.