From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from mimir.eigenstate.org ([206.124.132.107]) by ewsd; Sun Mar 22 16:52:25 EDT 2020
Received: from abbatoir.fios-router.home (pool-162-83-132-245.nycmny.fios.verizon.net [162.83.132.245])
	by mimir.eigenstate.org (OpenSMTPD) with ESMTPSA id 932c3cdd (TLSv1.2:ECDHE-RSA-AES256-SHA:256:NO);
	Sun, 22 Mar 2020 13:52:10 -0700 (PDT)
Message-ID: <3E7BBF2874236BB6E397A999AF9397E6@eigenstate.org>
To: plan9fullfrontal@qs.co.nz, 9front@9front.org
Subject: Re: [9front] Unsolicited patch for tclsh amd64
Date: Sun, 22 Mar 2020 13:52:09 -0700
From: ori@eigenstate.org
In-Reply-To: <e7c046a1-a282-881a-24a4-2c90432ec34f@qs.co.nz>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
List-ID: <9front.9front.org>
List-Help: <http://lists.9front.org>
X-Glyph: ➈
X-Bullshit: asynchronous polling CMS markup-scale base 

> I have issues with tclsh on amd64, this kludge got it running for me.
> 
> This may be helpful for others wanting to do a proper fix.
> 
> Test results show that there are differences (problems) with handling 
> stderr on exec and pipelines and other stuff.
> 
> It works enough for what I am doing.
> 
> NOTE: The addition of compile debug flag is necessary to stop tclsh from 
> bombing with invalid memory reference, remove and debug at you own leisure.
> 
> This fix is definitely not production ready.
> 
> Diff with cleanup performed under gnu diff.
> 
> diff -cr -x '*.acid' -x '*.tcl' -x '*.rc' -x '*.sh' 
> old/generic/tclExecute.c new/generic/tclExecute.c
> *** old/generic/tclExecute.c    2011-01-16 22:04:09.000000000 +1300
> --- new/generic/tclExecute.c    2020-03-09 23:06:36.000000000 +1300
> ***************
> *** 23,28 ****
> --- 23,30 ----
>    #include <math.h>
>    #include <float.h>
> 
> + typedef long long ptrdiffL_t;
> +

Use uintptr_t here; we should see if we can upstream that change.

>    /*
>     * Hack to determine whether we may expect IEEE floating point. The 
> hack is
>     * formally incorrect in that non-IEEE platforms might have the same 
> precision
> ***************
> *** 683,688 ****
> --- 685,693 ----
>    #endif /* TCL_COMPILE_STATS */
>    }
>    
> +
> + void snarf(Tcl_Interp *i,ExecStack *e) { return;}
> +
>    /*
> *----------------------------------------------------------------------
>     *
> ***************
> *** 705,711 ****
> *----------------------------------------------------------------------
>     */
> 
> ! #define TCL_STACK_INITIAL_SIZE 2000
> 
>    ExecEnv *
>    TclCreateExecEnv(
> --- 710,770 ----
> *----------------------------------------------------------------------
>     */
> 
> ! #define TCL_STACK_INITIAL_SIZE 8000
> ! void fnoop(void *) {}
> !
> ! typedef union {
> !     char *c;
> !     long long *w;
> ! } fudger;
> !
> ! void * fudge(void *ptr) {

Try updating -- I fixed this, we should be getting
back 16 byte aligned values. If it still doesn't
work, let me know.

changeset:   7610:3f0285bde94f
user:        Ori Bernstein <ori@eigenstate.org>
date:        Mon Mar 09 08:02:22 2020 -0700
summary:     fix alignment in ape malloc