[gentoo-dev] Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[gentoo-dev] Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Donnie Berkholz]

On 10:57 Thu 18 Oct     , Caleb Tennis (caleb) wrote:
> 1.1                  dev-libs/libevent/libevent-1.3e.ebuild
> 
> file : http://sources.gentoo.org/viewcvs.py/gentoo-x86/dev-libs/libevent/libevent-1.3e.ebuild?rev=1.1&view=markup
> plain: http://sources.gentoo.org/viewcvs.py/gentoo-x86/dev-libs/libevent/libevent-1.3e.ebuild?rev=1.1&content-type=text/plain

> pkg_postinst() {
> 	local ver="${prevver##*/}"
> 	if [[ -n "${ver}" ]] && [[ ${ver//[a-zA-Z-]} != "${PV//[a-zA-Z]}" ]] ; then

It isn't terribly likely to become an issue here, but it might be nicer 
to use [[:alpha:]] than [a-zA-Z].

> 		ewarn
> 		ewarn "You will need to run revdep-rebuild (included with app-portage/gentoolkit)"
> 		ewarn "to rebuild all packages that were built with libevent-1.0x."
> 		ewarn
> 		ewarn "Run the following to see which packages will be rebuilt:"
> 		ewarn "    revdep-rebuild --library ${ver}.so.1 -p"
> 		ewarn
> 		ewarn "If you are satisfied with the output, re-run without the '-p' to rebuild."
> 		ewarn
> 	fi
> 	unset prevver
> }

Thanks,
Donnie
-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Jan Kundrát]

[-- Attachment #1: Type: text/plain, Size: 320 bytes --]

Donnie Berkholz wrote:
>> 	if [[ -n "${ver}" ]] && [[ ${ver//[a-zA-Z-]} != "${PV//[a-zA-Z]}" ]] ; then
> 
> It isn't terribly likely to become an issue here, but it might be nicer 
> to use [[:alpha:]] than [a-zA-Z].

Is [[:alpha:]] locale-safe?

Cheers.
-jkt

-- 
cd /local/pub && more beer > /dev/mouth


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

[gentoo-dev] Re: Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Steve Long]

Jan Kundrát wrote:
> Donnie Berkholz wrote:
>>> if [[ -n "${ver}" ]] && [[ ${ver//[a-zA-Z-]} != "${PV//[a-zA-Z]}" ]] ;
>>> then
>> 
>> It isn't terribly likely to become an issue here, but it might be nicer
>> to use [[:alpha:]] than [a-zA-Z].
> 
> Is [[:alpha:]] locale-safe?
> 
Yes, all POSIX character classes listed here are:
http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap07.html


-- 
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Re: Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Jan Kundrát]

[-- Attachment #1: Type: text/plain, Size: 561 bytes --]

Steve Long wrote:
>> Is [[:alpha:]] locale-safe?
>>
> Yes, all POSIX character classes listed here are:
> http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap07.html

Thanks for a nice link. If I read section 7.3.1 correctly, [[:alpha:]]
always contains those letters, but might contain more, depending on the
locale. So it's probably very minor point, but as long as the script
runs with user-provided locale, one should be explicit here. Or am I
missing something here?

Cheers,
-jkt

-- 
cd /local/pub && more beer > /dev/mouth


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: [gentoo-dev] Re: Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Bo Ørsted Andresen]

[-- Attachment #1: Type: text/plain, Size: 895 bytes --]

On Wednesday 14 November 2007 01:33:54 Jan Kundrát wrote:
> Steve Long wrote:
> >> Is [[:alpha:]] locale-safe?
> >
> > Yes, all POSIX character classes listed here are:
> > http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap07.html
>
> Thanks for a nice link. If I read section 7.3.1 correctly, [[:alpha:]]
> always contains those letters, but might contain more, depending on the
> locale. So it's probably very minor point, but as long as the script
> runs with user-provided locale, one should be explicit here. Or am I
> missing something here?

If you think [a-zA-Z] is being explicit then you are clearly missing 
something. E.g. if you google for the estonian alphabet you'll see a lot of 
ansi letters are after z in that alphabet. So on estonian locales those 
letters won't be included in [a-z] but they will be included in [:alpha:].

-- 
Bo Andresen

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] Re: Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Jan Kundrát]

[-- Attachment #1: Type: text/plain, Size: 456 bytes --]

Bo Ørsted Andresen wrote:
> So on estonian locales those letters won't be included in [a-z] but
> they will be included in [:alpha:].

Actually that was exactly my point. If user had some funny locale that
has digits or dots in its [:alpha:], the ebuild wouldn't work as
desired, but it would rather strip valid stuff from the variable. But
that's probably pure sci-fi for now :).

Cheers,
-jkt

-- 
cd /local/pub && more beer > /dev/mouth


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: [gentoo-dev] Re: Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Bo Ørsted Andresen]

[-- Attachment #1: Type: text/plain, Size: 649 bytes --]

On Wednesday 14 November 2007 21:37:46 Jan Kundrát wrote:
> Bo Ørsted Andresen wrote:
> > So on estonian locales those letters won't be included in [a-z] but
> > they will be included in [:alpha:].
>
> Actually that was exactly my point. If user had some funny locale that
> has digits or dots in its [:alpha:], the ebuild wouldn't work as
> desired, but it would rather strip valid stuff from the variable. But
> that's probably pure sci-fi for now :).

[:alpha:] will never have digits or dots in it. It will, however, sometimes 
have funny letters in it. And [a-z] will not always have all ansi letters in 
it.

-- 
Bo Andresen

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[gentoo-dev] Re: Re: Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Steve Long]

Jan Kundrát wrote:

> Steve Long wrote:
>>> Is [[:alpha:]] locale-safe?
>>>
>> Yes, all POSIX character classes listed here are:
>> http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap07.html
> 
> Thanks for a nice link. If I read section 7.3.1 correctly, [[:alpha:]]
> always contains those letters, but might contain more, depending on the
> locale. So it's probably very minor point, but as long as the script
> runs with user-provided locale, one should be explicit here. Or am I
> missing something here?
> 
No, that's about the size of it-- if you you'd like to tie it to ASCII,
irrespective of locale, that's fair enough. It depends on what you're up to
(sorry, I don't have time to go digging through code to see what this
applies to) but in the /general/ case it's better to use locale-neutral
character-classes, since it makes scripts much more useful.

Setting LC_ALL=C temporarily is not a good idea, since it overrides
LC_CTYPE. The most common usage for that is for sort order; where that's
needed it's better to use LC_COLLATE. (man 7 locale)

Quick example showing why the double-bracket appears:
   if [[ ${v:0:1} != [[:alpha:]_] || $v = *[^[:alnum:]_]* ]]; then
      errMsg $"$v is not a valid identifier"
      return 1
   fi

Getting to know these is really helpful, imo, especially since they apply to
_all_ the utilities like tr, sed, grep, awk -- and ed ofc ;)

$"text" is for i18n in bash via gettext, akin to _"foo" in C. Can't say I've
used it yet, though :p


-- 
gentoo-dev@gentoo.org mailing list

[gentoo-dev] Re: Re: Re: [gentoo-commits] gentoo-x86 commit in dev-libs/libevent: ChangeLog libevent-1.3e.ebuild

[Steve Long]

Steve Long wrote:
> No, that's about the size of it-- if you you'd like to tie it to ASCII,
> irrespective of locale, that's fair enough.
> 
I had a feeling this statement was incorrect (what zlin said about a-z) so I
asked dalias in #bash who is knowledgeable about locale stuff:

<igli> dalias; have a q re: l10n. if we want to restrict allowed chars to
ASCII (for config) without affecting user's LC_CTYPE, what's the best way?
<dalias> igli, where do you want to restrict chars?
<igli> example i've seen is in checking a var
<dalias> checking a var to make sure it's all ascii?
<igli> yeah
<igli> [A-Za-z] isn't locale-safe is it?
<dalias> match it to a regex
<igli> hmm ok
<dalias> it's safe if LC_COLLATE=C
<igli>  ah thanks :D
<dalias> alternatively
<dalias> the most safe way is just to write all chars explicitly
<dalias> [ABCDEF...xyz]
<igli> hehe
<igli> ok nice one, i'll pass it on.
<dalias> anyway posix guarantees that the collation order in C locale is
ascii order
<dalias> even if the host's character encoding is something idiotic like
ebcdic

The odd bit (for me) is that [A-Za-z] is affected by LC_COLLATE not
LC_CTYPE, which makes sense if you think of it as a comparison. [[:class:]]
is ofc CTYPE.

So setting LC_COLLATE=C would appear to make sense, as long as you are never
dealing with user filenames, but only the portage tree (which seems a bit
yuck). A better general implementation might be a function that checks
against the explicit list, since that would have no implications for file
handling:

isAlphaASCII() {
for i; do
[[ $i = *[^ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyzxyz]* ]] \
        && return 1
done
return 0
}

I /guess/ in sh the test line would be:
case "$i" in *[!ABC..xyz]* ) return 1;; esac


-- 
gentoo-dev@gentoo.org mailing list