[Bug 69446] Potential race when generating gtypes (and more)

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=69446

Simon McVittie <simon.mcvittie at collabora.co.uk> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|telepathy-bugs at lists.freede |simon.mcvittie at collabora.co
                   |sktop.org                   |.uk
  Attachment #86851|0                           |1
        is obsolete|                            |

--- Comment #3 from Simon McVittie <simon.mcvittie at collabora.co.uk> ---
Created attachment 93055
  --> https://bugs.freedesktop.org/attachment.cgi?id=93055&action=edit
codegen: avoid generating more than one thing in the same  command

As with telepathy-glib commit 36c2a545c9, a rule like this:

    _gen/x.c _gen/x.h: prerequisites
        $(AM_V_GEN)x-generator

doesn't consider x.c and x.h together. Instead, it expands to two rules,
one to generate x.c and one to generate x.h, which happen to run the
same commands.

This means that in the worst case, you can end up running x-generator
twice in parallel, and they'll race with each other and overwrite or
delete each other's output.

Updated version of a patch from Ross Burton, taking into account that
we now generate a separate header for the gtk-doc. I use the gtk-doc
header as the one that "matters", because it's the last to be generated.

---

I'm proposing this patch even though I'm about to delete the code-generation
stuff, because (a) I just encountered it, and (b) I would like a `git revert`
of that deletion to bring us back to a correct state.

Not using Ross' patch directly because we now generate more files than we used
to.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.