X-Authentication-Warning:	delorie.com: mail set sender to djgpp-bounces using -f
X-Recipient:	djgpp AT delorie DOT com
X-Original-DKIM-Signature:	v=1; a=rsa-sha256; c=relaxed/simple; d=gmx.net;
	s=badeba3b8450; t=1608081619;
	bh=6JFL+pn0+Kifk6HWVULNnIn0TCyJN0s2jF+kPP1AAik=;
	h=X-UI-Sender-Class:Date:From:To:Subject:References:In-Reply-To;
	b=OquGWH5l7VX7cz5vDZTrx9ayOMWAG3EY48KE0jmLmchgaS/1ZKokyHsKAuf7AIaet
	nkmkmCaC8yufjRFCVZdIOLBAujGN4DS1fSqPtZv+kni4bD1ySU8nEm+iRwQvKJySOy
	hiw/SXJVDyS512gnsoHEOpk9ofQf1hTW6AqXOIW0=
X-UI-Sender-Class:	01bb95c1-4bf8-414a-932a-4f6e2808ef9c
Message-ID:	<5FD960D1.3010008@gmx.de>
Date:	Wed, 16 Dec 2020 02:20:17 +0100
From:	"Juan Manuel Guerrero (juan DOT guerrero AT gmx DOT de) [via djgpp AT delorie DOT com]" <djgpp AT delorie DOT com>
User-Agent:	Mozilla/5.0 (X11; U; Linux x86_64; de; rv:1.9.2.13) Gecko/20101206 SUSE/3.1.7 Thunderbird/3.1.7
MIME-Version:	1.0
To:	djgpp AT delorie DOT com
Subject:	Re: Difficulties compiling libc from repository using gcc 10.N.0
References:	<5FD1FEA1 DOT 7010304 AT gmx DOT de> <CAA2C=vB2FUmrC_iboB7Mp3Uu2hLoTxWB2ptEOkjc69Az3FZU1A AT mail DOT gmail DOT com> <5FD2C0B9 DOT 8080208 AT gmx DOT de> <CAA2C=vAj+1p9PNXSB3sJvrxFUn9SvgG6uHRSbxsx48+OsyYiwg AT mail DOT gmail DOT com> <CAA2C=vBU+YB_x_rkpmEc8G_8yccXK5MivRLMf5wGO2-jTwuWVQ AT mail DOT gmail DOT com>
In-Reply-To:	<CAA2C=vBU+YB_x_rkpmEc8G_8yccXK5MivRLMf5wGO2-jTwuWVQ@mail.gmail.com>
X-Provags-ID:	V03:K1:QzbHHunlR+F8jxCPAR6EeA86k3MPtW2/YndoynKjPp0Ga0TN6lm
	WShDwQiwSa88tmVNEBQA1Y+yBrkAn8UbqX7v3jFW3yc7iZyRzx1uNrtEa+oZ5fHBfaLEHCl
	VqLOOlwhCICXbEiFB3Rw6ZlWLHzm6SowYFNIn5sQ4KNhjdEu+XUCreuGhd0HeT3nQyc6Fyo
	u7a00nRtC2bYqAngB/toQ==
X-Spam-Flag:	NO
X-UI-Out-Filterresults:	notjunk:1;V03:K0:pXlWfpkJ2S0=:viDoJkxkKLwezhjMBgaJjk
	7IprhsSfJG02jU4XgfWd1ruiNN+Dk3xAQJQbDKl9y75RzJp+ff0ueTcqNpwS8EmyDE6X5/Jt4
	Fae42Qh35VP+febhSwnKsWcIQFtJBpKsT1jB1yFfzPsra/ASNa2G3ApaIJItvCWNRoPoBdQLV
	K6I99BjRk7HC960xFQD5xpPzyaURr7sSOuqP2jL2mgT222pH8HzoocQ6GQGLMfZ4JhZD1PMPH
	SB2jceh8moVtqOTDn/6EOAdDfaLzKZErtl4ninrhXL+hXO6yelZDV+pOXjMZyKfqAnc5Xy+cw
	IRKrggCabMz3WK7Cz2K3gqirDIeLFdUaGXE3nvYeG/0VbP3ebaf/Tz/XMcVYDxb1xZ3aZ/lQZ
	8V9gQL3uA/6OLfclnnB3ZqO1eQrz/k9twHVJCvcYQEBSEbafUIe0WB8KkE3WU0DT71mvtq7ER
	MGpZS4DTwRyspR8kklvXSaJ+VhiRQ88ylmcOked/S+5FaAeS5FHcZceUYDozraLtaGHtXZGJL
	GlZBHPaVSap25mGmmkoLoFIK7aRvMwIrJi31NaFM+c9mngIWH34l2zDalEYDW4/AY5r/CKSE1
	l5wmdrSEMPmlEg2mVdef526wzn7QBuLw3Tc5QRPwGQ5eVsasMIdYNWAkSMmp86nHvKKm9CN20
	ziyYgJ2vhdbmd5kaglfs/xWdJGD0uJRibFA3xCehaPo2nJXbisgAR9zi9lFYmz6A4l+i1SB+K
	j7RLZmHO4foV1N7vxZhoWeLVrZHDTrdnmGsdrwYavYTk1SVlC+id/MPcziKu0lOG0tiS34zU7
	TBNI5mzcBivCvkYxwUgy+zfFQuLYDFKwRlOBIKKzQ0DI3StvWG3u8nCPaCfXpB3haK6Qj+cm8
	vqLGsXFC0/UEtZVUpOHA==
Reply-To:	djgpp AT delorie DOT com

Am 13.12.2020 18:52, schrieb Ozkan Sezer (sezeroz AT gmail DOT com) [via djgpp AT delorie DOT com]:
> On 12/11/20, Ozkan Sezer<sezeroz AT gmail DOT com>  wrote:
>> On 12/11/20, Juan Manuel Guerrero (juan DOT guerrero AT gmx DOT de) [via djgpp AT delorie DOT com]<djgpp AT delorie DOT com>  wrote:
>>>> As far as I can see, this particular case is intending this
>>>> using a nasty shortcut?
>>> What is today a nasty shortcut was a cool one 30 years ago.
>>
>> No arguments there.
>>
>>>> Correct?  Any others like this?
>>> Looks ok to me.  Will do the job; alternatively we would have to disable
>>> that check.
>>
>> Attached a patch that applies to current CVS.
>>
>> However, the asm outputs differ.  Tried gcc-3.4.6 and gcc-8.2.0,
>> asm outputs from both are attached here (*_0.s for original and
>> *_1.s for patched.)  Also attached diffs of *.s for convenience.
>> Someone good with x86 asm should verify.
>
> Has anyone had a chance to look at this?

IMHO it makes not to much sense to compare the assembly code.  It is clear that
different code will produce different assembly output.  I do not even know how
the va_* macros are implemented and if they are provided by the libc code or by
the compiler as build-in features.  The only important thing is that the proposed
code is functional equivalent to the one that shall be substituted.  I have
tried to prove this with the small modification shown below and with a test
program that first reads the attribute then sets the read-only bit and writes
back the attribute.  I have done this with gcc345 and gcc1020 and in both case
it did work for me.  From my point of view the proposed fix is ok but it is only
opinion.

Regards,
Juan M. Guerrero





OFYI, the small check code:


diff -aprNU3 djgpp.orig/src/1.c djgpp/src/1.c
--- djgpp.orig/src/1.c	1970-01-01 00:00:00 +0000
+++ djgpp/src/1.c	2020-12-15 12:51:32 +0000
@@ -0,0 +1,20 @@
+#include <stdio.h>
+#include <io.h>
+
+int main(void)
+{
+  int f, m, rc;
+
+  f = 0;
+  rc = _chmod("1.c", f);
+  if (rc != -1)
+  {
+    m = rc | 0x00000001;
+    f = 1;
+    rc = _chmod("1.c", f, m);
+  }
+  if (rc == -1)
+    printf("failed");
+
+  return rc;
+}
diff -aprNU3 djgpp.orig/src/libc/dos/io/_chmod.c djgpp/src/libc/dos/io/_chmod.c
--- djgpp.orig/src/libc/dos/io/_chmod.c	2018-05-14 16:56:40 +0000
+++ djgpp/src/libc/dos/io/_chmod.c	2020-12-15 12:51:26 +0000
@@ -9,7 +9,9 @@
  #include <dpmi.h>
  #include <fcntl.h>
  #include <libc/dosio.h>
-
+#include <stdarg.h>
+#include <stdlib.h>
+
  int
  _chmod(const char *filename, int func, ...)
  {
@@ -25,7 +27,16 @@ _chmod(const char *filename, int func, .
      r.x.ax = 0x4300 + func;
    _put_path(filename);
    if (func == 1)
-    r.x.cx = *(&func + 1);		/* Value to set */
+  {
+    int v1, v2 = *(&func + 1);		/* Value to set */
+    va_list ap;
+    va_start (ap, func);
+    r.x.cx = v1 = va_arg(ap, int);		/* Value to set */
+    va_end (ap);
+
+    if (v1 != v2)
+      abort();
+  }
    r.x.dx = __tb_offset;
    r.x.ds = __tb_segment;
    __dpmi_int(0x21, &r);

- Raw text -