Bug#554167: Updating Mawk in Debian

Hi Yann,

wrote:

any news about updating Mawk with the last upstream version?

I don't think it can happen and be properly tested in time for wheezy.
The new upstream version has significant changes relative to the
packaged version and probably introduces some (minor or not)
regressions. (For example, the -W i option didn't work in Turkic
locales the last time I checked, because toupper('i') is not 'I'.)

What would be very helpful is code review. For example, collect
a batch of twenty or so patches (e.g., changes up to 1.3.3-20090705)
from [1] or straight from Thomas. File a bug that lists the patches,
their purpose, potential regressions, and how that potential can be
mitigated. Then I would be happy to help get those patches applied
in experimental.

Another way to help is to get the code history up to the present in a
readable state, to help that same effort. That basically means:

* improve the "rcs fast-export" tool[2]. All improvements are good. :)
Packaging it for Debian would be great because then we get a
bugtracker. If you can get the raw RCS files to test changes, that
might make this task easier.

* collect more changes (in RCS or git bundle form) and let me know
so I can update [1] to include the updated code.

Testing is of course also welcome.

Hope that helps,
Jonathan

[1] http://git.debian.org/?p=users/jrni...orical.git
[2] http://wok.oblomov.eu/tecnologia/rcs-fast-export/



To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact

On Mon, May 28, 2012 at 02:48:02PM -0500, Jonathan Nieder wrote:

Hi Yann,

wrote:

> any news about updating Mawk with the last upstream version?

I don't think it can happen and be properly tested in time for wheezy.
The new upstream version has significant changes relative to the
packaged version and probably introduces some (minor or not)
regressions. (For example, the -W i option didn't work in Turkic
locales the last time I checked, because toupper('i') is not 'I'.)

What would be very helpful is code review. For example, collect
a batch of twenty or so patches (e.g., changes up to 1.3.3-20090705)
from [1] or straight from Thomas. File a bug that lists the patches,
their purpose, potential regressions, and how that potential can be
mitigated. Then I would be happy to help get those patches applied
in experimental.

I don't recall seeing any comments from Steve on this.

Another way to help is to get the code history up to the present in a
readable state, to help that same effort. That basically means:

* improve the "rcs fast-export" tool[2]. All improvements are good. :)

It needs work. I commented on what I'd done a couple of months ago,
but saw no replies. (At the moment I'm actually working on mawk).

Packaging it for Debian would be great because then we get a
bugtracker. If you can get the raw RCS files to test changes, that
might make this task easier.

I have in mind creating a git-bundle using my fixed-up rcs-fast-export
(as an export-only...).

* collect more changes (in RCS or git bundle form) and let me know
so I can update [1] to include the updated code.

Testing is of course also welcome.

Hope that helps,
Jonathan

[1] http://git.debian.org/?p=users/jrni...orical.git
[2] http://wok.oblomov.eu/tecnologia/rcs-fast-export/

Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net






To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact

(culling cc list)
Thomas Dickey wrote:

I have in mind creating a git-bundle using my fixed-up rcs-fast-export
(as an export-only...).

Thanks. Please let me know when it's ready and I can publish it in
the same place as the earlier export.

Hope that helps,
Jonathan



To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact

On Mon, May 28, 2012 at 03:35:47PM -0500, Jonathan Nieder wrote:

(culling cc list)
Thomas Dickey wrote:

> I have in mind creating a git-bundle using my fixed-up rcs-fast-export
> (as an export-only...).

Thanks. Please let me know when it's ready and I can publish it in
the same place as the earlier export.

ok. Will do that after I get done with my current change...
It's related to this:
http://sources.redhat.com/ml/libc-a...00028.html

Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net






To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact

On Mon, May 28, 2012 at 03:35:47PM -0500, Jonathan Nieder wrote:

(culling cc list)
Thomas Dickey wrote:

> I have in mind creating a git-bundle using my fixed-up rcs-fast-export
> (as an export-only...).

Thanks. Please let me know when it's ready and I can publish it in
the same place as the earlier export.

hmm. I think it's more complicated than that. I've created a git-bundle,
but it requires some discussion. Here's a recap:

+ I spent a chunk of time in January working with rcs-fast-export,
and setup a script to fixup the email addresses, to make the logs
work as expected. I modified rcs-fast-export to accept an option
which makes it resolve the RCS identifiers properly. (You may
recall from earlier discussion that the identifiers _stored_ in
the archive are from the previous data, rather than the actual
revision). That's slower to execute, of course.

+ In January I also worked (without much success) to develop a
script to splice my changes onto a copy of the git bundle that
you had published. There is more than one apparent problem
(aside from my own lack of git experience). With or without
the corrected RCS identifiers, the process fails due to merge
problems. I recall that this only gets about 60% through the
import. (I can of course provide you with the corresponding
scripts, and we can discuss better approaches to the problem).

+ Not having progress on git (and having agreed to provide that),
the dependent activities were stalled. However, I'm back on
mawk for a few weeks. As I indicated, I noticed an interesting
problem using mawk in a case where it's a matter of implementing
a gawk-extension.

I can see that extension would be a distraction (so I put it
aside for the moment).

+ On a related thread, it was suggested that I get mawk into
savannah for using bug-tracking. When I investigated that,
I noticed that they rather insist on having each file marked
with copyrights. Both Mike and I had not done that as rigorously
as I think savannah wants. So... I made that change on top of
my 20101207 tag, calling that 20101210 (and kept the timestamps
on that date to avoid confusing git further). The tag is "old"
of course since it's solely done to mark copyrights as of that
date, and to avoid confusion with other changes that I have in
mind.

I put the mawk bundle for that in
ftp://invisible-island.net/GIT/
and uploaded that version as a new release on my webpage.

I think those are enough pieces in place that I can followup on
savannah; but have an idea that we need some discussion on how
to proceed for your idea of providing patches based on git.
(It would be nice if there were some way to record the transformation
from my rcs/git archive into yours so that I could generate a better
export).

Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net






To UNSUBSCRIBE, email to
with a subject of "unsubscribe". Trouble? Contact