This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: ChangeLog fixes


Kevin Buettner wrote:
> 
> For real historians, I would imagine that it's incredibly tricky to
> figure out whether such corrections are a rewriting of history in the
> pejorative sense or are truly accurate corrections.  Our historical
> texts are the ChangeLog entries, and for these, we (or some of us
> anyway) have the luxury of being able to check the veracity of the
> ChangeLog entries by means of the CVS repositories.

I have a little familiarity with what real historians do, and this
process is called "textual editing".  For instance, most of the Greek
and Roman literature exists today only as manuscript copies made by monks,
the oldest of those still a thousand years separated from the originals.
In the worst cases, a monk might not know the language of the text he
was copying and make serious nonsensical mistakes, then the next
monk would try to fix them while making his copy, and so today we end
up with a dozen manuscripts, all inconsistent with each other.  Even
worse, in some cases the mistakes completely change the sense of a
passage, and there are learned papers debating which interpretation
is more plausible, based on context ("internal evidence") or other
documents, archaeology, etc ("external evidence").

> Now to specifics...  The bulk of Michael's ChangeLog fixes concern the
> use of the ``*'' wildcard.  He left the original author's comment the
> same; all he did was expand the ``*'' to list all of those files which
> were actually affected.  As I understand it, he verified the accuracy
> of his rewrites by checking against CVS.

This seems like a reasonable thing, but beware of ChangeLogs that
predate CVS.  Since there are many secret files mentioned in the
internal RH repositories, of necessity he must have done that
additional edit of wildcard expansions as well.  So I think the
change is fine, but as the GNU coding standard mentions, the ChangeLog
is still just an approximation to what actually happened.

Because of the likelihood of compounding errors by rewriting old
ChangeLogs, in general I would recommend resisting the temptation
to try to improve them.  (I've had to hold myself back many times.)

Stan


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]