From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <002301c50e8a$56392af0$7c18dfd5@kilgore>
From: "boyd, rounin" <boyd@insultant.net>
To: "Fans of the OS Plan 9 from Bell Labs" <9fans@cse.psu.edu>
References: <41F879E200020C44@stateless1.tiscali.cz>
Subject: Re: [9fans] 8c (pcc) bug??
Date: Wed,  9 Feb 2005 10:32:55 +0100
MIME-Version: 1.0
Content-Type: text/plain; format=flowed; charset="ISO-8859-2";
	reply-type=original
Content-Transfer-Encoding: 7bit
Topicbox-Message-UUID: 4429898c-eace-11e9-9e20-41e7f4b1d025

this is one nasty bug.

the problem is the // and /* comments are parsed twice:

    1) in a #define
    2) in an ordinary code block

the assumption that /sys/src/cmd/cc/macbody makes [case #1] is that
a #define consists of a:

    - name
    - optional list of args
    - list of names (possibly arguments)
    - line continuations
    - comments
    - characters

should you have the misfortune to:

    #define S "//...

or

    #define  S "/*...

you get:

    #define S "

and you are in a "world of pain".

to my mind it should be taught about strings, but this is somewhat non-trivial.

well, maybe not.  it needs to know about comments to skip arguments that
maybe are hidden away in comment.  ahh, this give me an idea.

/sys/src/cmd/cc/lex.c [case 2] goes to some effort to lex up strings and pass
them up to yacc.  it uses esschar() which decodes chars & Runes in char and
string constants, Rune by Rune.

i'm not a language lawyer, nor do i have copies of the numerous standards,
but do i do believe that:

    #define S "//foo"

should give me:

    "//foo"

hmm, yeah, i think i have a fix.
--
MGRS 31U DQ 52572 12604