||i wrote:
 |||After about ten months of development i have merged into [master] the
 |||current state of affairs.

So it is driving me insane, but i should have waited a bit, Sun
users have to, others should use what is now on [master], instead
of using [v14.9.0-pre1]!  There will be no -pre2 due to this,
though.  Sorry for that inconvenience.

It turned out we have a general problem on at least Sun Solaris,
where +00C1 a.k.a. \u00C1 will be happily iconv(3)d to "?", i.e.,
i can happily iconv_open(3) from UTF-8 (128,172 characters) to ISO
646 a.k.a. US-ASCII (128 characters).

The thing is, i was happy to be able to support any \u or \U
sequence even in non-Unicode locales as long as the character is
actually representable in the target locale, counting on iconv(3)
to fail (yes, in the meanwhile i know that POSIX doesn't specify
such an error, but, i mean, come on: isn't that ...).
But it seems that was no good idea, the test fails under Sun
Solaris because of this.  So, this part of the shell quoting
mechanism has to be revisited, until at least S-CText is usable
and we can simply and regular check whether conversion would
succeed.  Pain.

Don't worry it will all end soon 
The crack of doom is coming soon.



Reply via email to