2000-11-04  Not Zed  <NotZed@HelixCode.com>

        * camel-mime-part-utils.c
        (camel_mime_part_construct_content_from_parser): Set the
        preface/postface from the parser into the multipart object.

        * camel-multipart.c (camel_multipart_set_postface): Function to
        set the postface text on a multipart.
        (camel_multipart_set_preface): Similarly for preface text.

        * camel-mime-parser.c (folder_scan_content): If we scan until a
        boundary, then we do not include the \n that starts the boundary
        line in the content.
        (struct _header_scan_stack): Added a ByteArray to store the
        multipart pre/post-text as we're scanning.
        (folder_pull_part): Free pre/posttext if they are allocated.
        (folder_scan_step): Build into the pre/posttext arrays as we
        encounter data.
        (camel_mime_parser_preface): REturn the multipart preface text, if
        there is any scanned.
        (camel_mime_parser_postface): Likewise for postface text.
        (byte_array_to_string): helper function for above.

        * providers/mbox/camel-mbox-folder.c (mbox_append_message): Change
        the from line to be "\nFrom ..." always, so no need to
        check/append a \n to messages.
        (mbox_append_message): Open the output stream with append mode
        [assuming this is more efficient than seeking to the end]
        And dont prepend \n  on the From line if its the first in the
        mbox.
        (mbox_append_message): Pass the offset of the real start of the
        "From " line when we perform the update (which may != 'seek')

        * camel-mime-filter-charset.c (complete): Removed the terminating
        NUL 'fix'.

        * camel-stream-filter.c (do_read): Added some debug.
        (do_flush): And here.
        (do_write): And here too.
        (do_write): ARGH!!! ARGH!  Ok, so the filter stream was writing a
        different number of bytes than the requester was asking it to
        write (because of filtering, of course!).  So instead of returning
        the true number of written bytes, we'll return what they asked us
        to write - unless there is an error in which case we return -1.

        * camel-mime-utils.c (base64_encode_close): Sigh, forgot to make
        it unsigned.  I think this is actually a gcc bug as (48 >> 2)
        somehow ended up negative, when it obviously should not, even if
        the data load was signed.

svn path=/trunk/; revision=6388