[GRASS-dev] svn access

[Radim Blazek]

On Fri, May 15, 2009 at 11:40 AM, Hamish <hamish_b at yahoo.com> wrote:
>
> Radim wrote:
>> OK, patch attached, fix for QGIS crash.
>
> thanks,
>
>> I know, I should create a new ticket and then make a
>> comment with the patch attached, right?
>
> Right, as it's the only sane way to deal with the extreme information
> overload to people-power ratio without sucking up all of our free time.
> Also it gives the patch some air and makes it easy to help teach the
> rest of us not to make the same mistakes again. FWIW I was a bit skeptical
> about doing everything in trac to begin with but have really come around
> to appreciate its usefulness. The grand idea is to minimize overheads...
>
>
> e.g. Is there a qgis ticket for this?

Probably no. I dont know.

> ie how to trigger the crash and why does it happen?

Array was not zero terminated.

> is it only vs. grass-trunk? i.e. how can I test
> the patch & know if it needs backporting/expanding/whatever?

The same problem in all versions.

> should the real fix be "alloc += GMAPSET_MAX"?

I dont think so, dynamic allocation is used.

> or to modify/fix the header comments to say that the pointer will
> be null if the location dir can't be opened or no mapsets are found?

No, it says 'zero terminated array' so it should not be changed now so
that it can return NULL, I think.

> (and then submit a patch to qgis)
> [now: "\return pointer to zero terminated array of available mapsets."]
>
> and was the final null termination missing?;

Yes.

> which, if so, would need to be backported to 6.4/6.5 as well?

Probably. g.mapsets also calls that function but it is not big problem
if it crashes at the end instead of normal exit.

It seems that until now allocated space was inited to 0 also by
malloc, is it possible?

Radim


> ... I'm just trying to understand it, and honestly the trac stuff really
> does help with that because it helps focus the issue + automatic URLs ..
>
>
> Hamish wrote:
>> > a post to the grass-psc list w.r.t. RFC2 is needed,
> ....
>> > (we all had to do it; you are already listed in
>> > trunk/contributors.csv as a committer for whatever that means)
>
> in hindsight & looking at rfc1 again, it means that you were already
> "voted in" for write-access long ago under the mechanism of the day,
> so you are not a new devel and there is no need for all that; all that
> is needed for activation is a post to the psc list saying you agree
> with rfc2 (same as other cvs-converted devels had to).
> This formality was the "cost" we paid to join with OSGeo. It is the terms
> of legal incorporation in most places that state that these (often
> redundant) procedures have to be in place, ie why we had to bother with
> the RFCs at all.
>
>
> regards,
> Hamish
>
>
>
>
>
>