[patch, libgfortran] [4.3/4.4/4.5 Regression] I/O: Problems reading partial records in formatted direct access files

[Jerry DeLisle]

On 11/19/2009 12:34 AM, Tobias Burnus wrote:
> On 11/19/2009 05:47 AM, Jerry DeLisle wrote:
>>> OK.  Are we still "supporting" 4.3?
>>>
>> Not too sure on 4.3.  I am not even sure the patch would be the same
>> there, but I will check.
>
> Well, the official GCC policy is to support the current development
> version (i.e. 4.5), the last stable release series (i.e. 4.4.x) and the
> previous stable release series (i.e. 4.3.x). Support means fixes for
> reported regressions will be considered to be applied also to stable
> branches.
>
> Whether a given bug fix will be backported to an earlier release or not,
> depends on the maintainers (i.e. on us). In tendency, regressions have a
> higher priority than other bugs, wrong-code bugs a higher priority than
> ICE-on-valid or rejects-valid PRs. And patches with few lines which are
> obvious are better suited for backporting than long, complicated patches.
>
> Regarding 4.3 and this regression: It is a wrong-code PR with a simple
> 2-line patch - and it is a regression. On the other hand, it is an old
> bug (working in 4.1, failing in 4.2/4.3/4.4/4.5) and it has not been
> found before - additionally, 4.3.x will loose the supported status as
> soon as 4.5. is released, which will happen in few months (personal
> guess: around end of February).
>
> Personally, I would do the backport to 4.3, but I am also fine with
> 4.4/4.5 only.
>
> Tobias
>
The patch does apply to 4.3 manually since it is pre Janne's rework of the I/O.

I will go ahead and apply it a few days to 4.4 and 4.3.  I do want to see if any 
issues pop up with it on trunk.

Thanks for feedback,

Jerry