[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: RFC: mawk (not maintainer, updated package)



Dear Charles,

Charles Plessy wrote:
> Le Sun, Dec 13, 2009 at 12:04:17AM -0600, Jonathan Nieder a écrit :

>> My motivation was that the Linux kernel source started requiring an
>> awk that can understand [[:lower:]].  I imagine others will start
>> running into the same thing sooner or later: mawk 1.3.3 does not
>> support POSIX regular expressions very well, making it a poor default
>> awk.  The updated version is far better in that respect.

> I think that for packages of priority higher than optional, it is much safer to
> discuss their situation on -devel and decide what action to take there instead
> of proposing an NMU on -mentors. These packages are installed on most Debian
> systems, so uncoordinated uploads can be very disruptive.

This is a very good point.  I should have at least set the
distribution of the demonstration package to experimental.  Anyway, I
would like to wait a little for the maintainer to look at it, because
in a case like this (upgrading to a new upstream version with no
obvious backward incompatibilities), it should not be necessary to ask
others to help make a decision.

I decided to ask about the package on -mentors for two reasons, only
one of which I articulated clearly.  One reason was to learn how to
make sure something gets done; you have suggested a very reasonable
answer here (discuss the situation on -devel).  The other reason was to
get feedback on the changes I made.

If anyone wants to look, probably the easiest way is to look over the
changes in git. [1]  The main change is in the single commit b209e5a8
(Switch to Thomas Dickey’s new upstream version) [2].  Everything
since then is about making the build a little more comfortable: clean
up debian/rules, address warnings from gcc, allow build with byacc
instead of bison as yacc.  Thoughts on any of those things, especially
the debian/rules, would be welcome.

> In the case of the problem you mentionned above, maybe it can be solved by
> using another awk alternative?

gawk works.

original-awk might be worth considering as awk for the base system.
Will follow-up on -devel tomorrow.

Both original-awk and the updated mawk deal even worse with Linux than
the mawk 1.3.3.  (One can test with

  awk -f arch/x86/tools/gen-insn-attr-x86.awk arch/x86/lib/x86-opcode-map.txt

with a mainline kernel.)  mawk 1.3.3 runs afoul of the sanity check in
check_awk_implement(), producing a "Please try to use gawk." error.
original-awk and recent mawk use a different iteration order than gawk
for

 for (i in opnd)

in convert_operands, which makes the function trip up on input line 240:

 c8: ENTER Iw,Ib

because they produce the Iw operand before Ib.  Looks like this script
has only been tested with gawk. :-)

> Anyway, thank you for the care you have for core packages. Please do not
> be shy, and discuss them on a core mailign list :)

Thanks.  The more I think about it, the better advice this seems.

Good night,
Jonathan

[1] git.debian.org:/git/collab-maint/mawk.git
[2] http://git.debian.org/?p=collab-maint/mawk.git;a=commit;h=b209e5a


Reply to: