stdcxx-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Martin Sebor <>
Subject Re:
Date Thu, 17 Aug 2006 20:46:41 GMT
Liviu Nicoara wrote:
> Martin Sebor wrote:
>>Farid Zaripov wrote:
>>>  I can't understand what does the line "test -z "`echo $it | grep .cm`"
>>>;" in function below?
>>My guess is that whoever wrote it (cough, Liviu, cough) meant
>>a literal period [...]
> I can only assume that in some incarnation, a certain locale utility,
> outputs *.cm filenames when invoked with -m. GNU's doesn't, and it looks
> like neither does MS'.

Tru64 locale does that:

$ uname -sr && locale -m
OSF1 V5.1

But I'm not sure it matters. AFAICT, locale -m outputs the names
of the installed charmaps which are usually the same as the names
of the character description files (our locale utility behaves
that way). Only one of our character description files has a .cm
suffix (, formerly called We never had any
other .cm files, so I also have no idea why the test is looking
for .cm. It also seems that the test should do more than just
look for the .cm extension in the output. If the -m option isn't
recognized by the utility and produces an error message on stderr
(and even causes the utility to exit with a non-zero status), the
test will pass. That's definitely not what we want.

I think we should get rid of this script and replace it with
a regular test written in C that fully exercises the behavior of
the utility. That way we'll have to maintain just one version of
the test across all platforms instead of two (one for UNIX and
one for Windows). What do you think, Farid?


View raw message