CVS commit: pkgsrc/devel/guile-www

["Greg Troxel" <gdt%netbsd.org@localhost>]

Module Name:    pkgsrc
Committed By:   gdt
Date:           Wed Nov  7 00:50:06 UTC 2012

Modified Files:
        pkgsrc/devel/guile-www: Makefile PLIST distinfo

Log Message:
Update to 2.35.

- 2.35 | 2012-11-06

  - support for "make installcheck"

        This does "make check" at its core, so the same env vars apply.
        Obviously, you should do "make install" first for sane results.

  - ‘receive-response’ spec takes thunk for ‘custom’ handling

        This means that instead of using:

        (list 'custom (values P1 P2 P3 P4))

        to construct a ‘receive-response’ spec, you can use:

        (list 'custom (lambda () (values P1 P2 P3 P4)))

        In fact, this is now the preferred way, as support for using
        ‘values’ directly is now deemed a misfeature and marked for
        eventual removal (see below).

  - tested against Guile 2.0.6

        There are some deprecation warnings, however.

  - MARKED FOR REMOVAL

    - 2013-05-15 support for values in ‘receive-response’ spec

        Out of an explicit ‘call-with-values’ context, Guile 2.x will
        silently discard all values but the first if the continuation
        expects one value only (the common case).

        (cons 0 (values 1 2)) => (0 . 1)

        Thus, the ‘receive-response’ spec constructed by:

        (list 'custom (values P1 P2 P3 P4))

        fails to preserve P2, P3, P4 (w/ Guile 2.x).  See above for
        thunk support, the future-proof way to specify ‘custom’ procs.

  - bootstrap tools upgrade
    - GNU Autoconf 2.69
    - GNU Automake 1.12.4
    - Guile-BAUX 20121026.1208.c21074a

- 2.34 | 2012-03-29

  - bug fix: unbotch ‘upath->filename-proc’
  - bug fix: unbotch ‘log-http-response-proc’

        In Guile-WWW 2.33 (2012-02-29), these procs were changed to use
        (ice-9 optargs), but the change was botched.

  - bug fix: ‘(www http) http:post-form’ reserves ‘#\&’ and ‘#\=’

        That is, in the absense of upload specs, the urlencoding for the
        key names and values now also includes ampersand and equal sign,
        as those used to delimit pairs and separate key and value within
        a pair, respectively.

  - bug fix: ‘url-coding:encode’ outputs two hex digits for 0-15

        Previously, octets 0 through 15 were output as "%0" through
        "%f", respectively.  Now, they are "%00" through "%0f".

  - portability fix: big dishing loop handles #f ‘sockaddr:path’

        For Unix domain sockets, some Guile versions have ‘sockaddr:path’
        sometimes return the empty string, others ‘#f’.

  - ‘(www url) url:parse’ can handle IPv6 addresses

        Handling doesn't imply validation, however.  For example:
        (url:host (url:parse "http://fool@[::ffff:bogus.0.2.1]:42/foo";))
        => "[::ffff:bogus.0.2.1]"

  - ‘(... answer) hqf<-upath’ more efficient

        The char-set computation has been moved to load-time.

  - ‘(... cookies) simple-parse-cookies’ takes optional arg SEP

        By default, tokens are ‘#\,’ (comma)-separated.  You can use SEP
        to specify another character, e.g., ‘#\;’ (semicolon).

  - new ‘(www cgi) cgi:init’ option: cookies-split-on-semicolon

        If the server provides the CGI script with a "Netscape style"
        (i.e., old and deprecated) set of cookies, you can use this
        option to split them.

  - ‘(www cgi) cgi:getenv’ recognizes some more "idiomatic" names

        The following keys now are recognized by ‘cgi:getenv’ (that is,
        they no longer signal "unrecognized key" error): ‘server-name’,
        ‘server-software’, ‘server-protocol’, ‘auth-type’, ‘http-accept’.
        Also, the manual now marks those keys that either are aliases,
        or return parts of the associated value, of these idiomatic names.

  - new (... parse-request) proc: receive-request

        This uses the same underlying machinery as ‘receive-response’
        from (www http), and thus sports the same features.

        Additionally, there is a new ‘request’ type with five accessor
        procs: request-{method,upath,protocol-version,headers,body} and
        the new predicate ‘request?’.

  - new (... log) proc: string<-sockaddr

        This formats AF_UNIX or AF_INET addresses (including port).

  - new (... answer) proc: compose-response

        This is a stripped down variant of ‘mouthpiece’.  It uses the
        same underlying machinery as ‘send-request’ from (www http) and
        thus sports the same features.

  - MARKED FOR REMOVAL

    - 2012-12-31 module (www main)

    - 2012-12-31 (www url) procs
      - url:address
      - url:unknown

    - 2013-02-28 (... parse-request) procs
      - read-first-line
      - read-headers
      - skip-headers
      - read-body

    - 2013-02-28 (www http) procs
      - http:head
      - http:get

  - test suite expanded

        Previously, there was just one test.  Now, there are several,
        and they are in directory v/ (for "validation", "verification",
        "vouchsafed", and so on).

  - doc license upgraded to GNU FDL 1.3

  - maintenance uses Guile-BAUX 20120309.1509.1c4bb92

- 2.33 | 2012-02-29

  - distribution now .tar.xz

        If you have GNU tar, you can use "tar xf" and it will DTRT.
        If not, you can use "xz -dc TARBALL | tar xf -" to unpack.

  - configure no longer checks for some widely-available modules

        The modules ‘(ice-9 and-let-star)’, ‘(ice-9 rdelim)’ and
        ‘(ice-9 regex)’ have been widely-available since Guile 1.4.x.

  - configure uses SNUGGLE for cross-Guile portability

        SNUGGLE is Free Software: <http://www.gnuvola.org/software/>.
        It provides ‘build-aux/snuggle.m4’, which is more amicable than
        the various ‘guile.m4’ in the wild.

  - dropped ‘(www server-utils answer) string<-header-components’

        This was pre-announced below in Guile-WWW 2.28 (2010-06-22).

  - changes to (www http)

    - bug fix: ‘http:request’ sends message body as-is

        This procedure used to erroneously send ‘CRLF’ after each string
        specified in BODY.  Now, it sends BODY only.

    - ‘http:status-ok?’ also accepts numeric arg

        E.g., ‘(http:status-ok? 200)’ and ‘(http:status-ok? "200")’
        both return ‘#t’.

    - ‘http:request’ also accepts a symbolic method

        Of course, a string is still acceptable.  Additionally, it is
        now an error if ‘method’ is neither string nor symbol.

    - new procs: ‘send-request’, ‘receive-response’

        These give (www http) a more asynchronous feel; it is now
        possible to implement pipelining, connection re-use, etc.
        Support for "chunked" transfer encoding and other HTTP 1.1 bits
        is built-in, although you can explicitly downgrade to 1.0 if you
        wish.

        Other features: (optional) caller-directed i/o to reduce data
        motion; support for ‘u8vector’ body (see example in the docs);
        support for custom body type; caller-specifiable header-name
        normalization; procedure names w/o "http:" prefix :-D.

    - message field types now documented

        The message object returned by ‘receive-response’ has a numeric
        status-code and header names normalized by the caller.

        In contrast, for backward compatability, the message object
        returned by ‘http:request’ has a string status-code and
        lower-case header names.

        (Header names are symbolic in any case.)

    - new fluid: protocol-version

        This affects how ‘http:request’ calls ‘send-request’ internally.
        It also impacts ‘http:get’, ‘http:head’, and ‘http:post-form’,
        which call ‘http:request’.

        The default value is ‘(1 . 0)’, i.e., HTTP 1.0.  The idea is to
        support backward compatability out of the box, but allow users
        to experimentally do:

          (set-fluid! protocol-version '(1 . 1))

        without having to modify their current code.  (New code
        should use ‘send-request’ and ‘receive-response’ directly.)
        This is a transitional feature of Guile-WWW 2.x; 3.x will
        drop ‘http:request’ (et al) along with the need for this
        fluid, entirely.

  - (www cgi) recognizes ‘application/x-www-form-urlencoded’ better

        Before, Content-Type recognition was over-constrained, failing
        to handle parameters (usually of the form "; charset=...").
        Things are now more relaxed.

  - documentation now explicitly UTF-8

        This has no immediate impact, but if you regenerate the .info,
        it may be possible that the makeinfo you use will generate nicer
        quotes, e.g., ‘foo’ instead of `foo'.  Time will tell.

  - maintenance: Autoconf 2.68, Automake 1.11.3, Guile-BAUX 20111227


To generate a diff of this commit:
cvs rdiff -u -r1.30 -r1.31 pkgsrc/devel/guile-www/Makefile
cvs rdiff -u -r1.11 -r1.12 pkgsrc/devel/guile-www/PLIST \
    pkgsrc/devel/guile-www/distinfo

Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.