NetBSD Problem Report #49408

From www@NetBSD.org  Fri Nov 21 12:03:01 2014
Return-Path: <www@NetBSD.org>
Received: from mail.netbsd.org (mail.netbsd.org [149.20.53.66])
	(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits))
	(Client CN "mail.netbsd.org", Issuer "Postmaster NetBSD.org" (verified OK))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 48C97A66A6
	for <gnats-bugs@gnats.NetBSD.org>; Fri, 21 Nov 2014 12:03:01 +0000 (UTC)
Message-Id: <20141121120259.DB7C3A66A6@mollari.NetBSD.org>
Date: Fri, 21 Nov 2014 12:02:59 +0000 (UTC)
From: jmcneill@invisible.ca
Reply-To: jmcneill@invisible.ca
To: gnats-bugs@NetBSD.org
Subject: XftGlyphSpecRender crash with gcc 4.8.4 on earmv7hf
X-Send-Pr-Version: www-1.0

>Number:         49408
>Category:       toolchain
>Synopsis:       XftGlyphSpecRender crash with gcc 4.8.4 on earmv7hf
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    toolchain-manager
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Nov 21 12:05:00 +0000 2014
>Closed-Date:    Sat Jul 06 17:01:27 +0000 2019
>Last-Modified:  Sat Jul 06 17:01:27 +0000 2019
>Originator:     Jared McNeill
>Release:        7.99.1
>Organization:
>Environment:
NetBSD a31 7.99.1 NetBSD 7.99.1 (HUMMINGBIRD_A31) #381: Thu Nov 20 09:20:40 AST 2014  jmcneill@megatron.local:/Users/jmcneill/branches/HEAD/src/sys/arch/evbarm/compile/obj/HUMMINGBIRD_A31 evbarm

>Description:
Unless you build xftrender.c with -O0, libXft crashes like this:

  Memory fault (core dumped) 

#0  0x400b8ed0 in XftGlyphSpecRender () from /home/jmcneill/libXft.so.3
#1  0x400b7fe4 in XftDrawGlyphSpec () from /home/jmcneill/libXft.so.3
#2  0x400b8268 in XftDrawCharSpec () from /home/jmcneill/libXft.so.3
#3  0x0004a5a4 in xtermXftDrawString.part.0 ()
#4  0x0004a67c in drawClippedXftString ()
#5  0x0005032c in drawXtermText ()
#6  0x000227f8 in ShowCursor ()
#7  0x0002a6c0 in VTparse ()
#8  0x0002a858 in VTRun ()
#9  0x0005abe8 in main ()

>How-To-Repeat:
Edit src/external/mit/xorg/lib/libXft/Makefile and remove the -O0 hack, rebuild libXft, and run "xterm -fa monospace".
>Fix:

>Release-Note:

>Audit-Trail:
From: Martin Husemann <martin@duskware.de>
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: xsrc/49408: XftGlyphSpecRender crash with gcc 4.8.4 on earmv7hf
Date: Sat, 22 Nov 2014 21:10:01 +0100

 Actually looks like a compiler bug to me:

 (gdb) x/16i $pc-32
    0x400b8eb0 <XftGlyphSpecRender+1012>:        str     r5, [r4, #448]  ; 0x1c0
    0x400b8eb4 <XftGlyphSpecRender+1016>:        ldr     r5, [r4, #448]  ; 0x1c0
    0x400b8eb8 <XftGlyphSpecRender+1020>:        mla     r8, r12, r10, r8
    0x400b8ebc <XftGlyphSpecRender+1024>:        str     r8, [r4, #408]  ; 0x198
    0x400b8ec0 <XftGlyphSpecRender+1028>:        ldr     r8, [r4, #444]  ; 0x1bc
    0x400b8ec4 <XftGlyphSpecRender+1032>:        add     r4, r8, r5, lsl #2
    0x400b8ec8 <XftGlyphSpecRender+1036>:        
     sub r5, r11, #8192  ; 0x2000
    0x400b8ecc <XftGlyphSpecRender+1040>:        ldr     r5, [r5, #448]  ; 0x1c0
 => 0x400b8ed0 <XftGlyphSpecRender+1044>:        ldr     r8, [r5, #444]  ; 0x1bc
    0x400b8ed4 <XftGlyphSpecRender+1048>:        str     r7, [r8, r5, lsl #2]
    0x400b8ed8 <XftGlyphSpecRender+1052>:        
     sub r8, r11, #8192  ; 0x2000
    0x400b8edc <XftGlyphSpecRender+1056>:        mov     r5, #1
    0x400b8ee0 <XftGlyphSpecRender+1060>:        mov     r7, r8
    0x400b8ee4 <XftGlyphSpecRender+1064>:        str     r5, [r8, #448]  ; 0x1c0
    0x400b8ee8 <XftGlyphSpecRender+1068>:        ldr     r8, [r8, #408]  ; 0x198
    0x400b8eec <XftGlyphSpecRender+1072>:        str     r8, [r4, #4]

 and it dies because r5 is NULL:

 (gdb) info reg
 r0             0x3      3
 r1             0x40b06e60       1085304416
 r2             0x7fffa850       2147461200
 r3             0x0      0
 r4             0x7fff8dc8       2147454408
 r5             0x0      0
 r6             0x4094d000       1083494400
 r7             0x2600011        39845905
 r8             0x7fff8dc8       2147454408
 r9             0x0      0
 r10            0x1      1
 r11            0x7fffa7f4       2147461108
 r12            0x0      0
 sp             0x7fff8968       0x7fff8968
 lr             0x400b8b88       1074498440
 pc             0x400b8ed0       0x400b8ed0 <XftGlyphSpecRender+1044>
 cpsr           0x600e0210       1611530768

 This happens here:

 (gdb) list
 294                     if (n)
 295                     {
 296                         elts[nelt].nchars = n;
 297                         nelt++;
 298                     }

 299                     elts[nelt].glyphset = font->glyphset;
   ...... crash here

 300                     elts[nelt].chars = char8 + size * j;
 301                     elts[nelt].xOff = glyphs[i].x - x;
 302                     elts[nelt].yOff = glyphs[i].y - y;
 303                     x = glyphs[i].x;

 and a few values, according to gcc, are:

 (gdb) p elts
 $9 = (XGlyphElt8 *) 0x7fff8dc8
 (gdb) p nelt
 $10 = 0
 (gdb) p font
 $11 = (XftFontInt *) 0x4094d000
 (gdb) p &font->glyphset
 $12 = (GlyphSet *) 0x4094d080
 (gdb) p &elts[nelt]
 $13 = (XGlyphElt8 *) 0x7fff8dc8
 (gdb) p &elts[nelt].glyphset
 $14 = (GlyphSet *) 0x7fff8dc8


 Martin

Responsible-Changed-From-To: xsrc-manager->toolchain-manager
Responsible-Changed-By: jmcneill@NetBSD.org
Responsible-Changed-When: Sun, 23 Nov 2014 23:55:00 +0000
Responsible-Changed-Why:
Toolchain problem.


State-Changed-From-To: open->feedback
State-Changed-By: maya@NetBSD.org
State-Changed-When: Tue, 25 Dec 2018 13:44:16 +0000
State-Changed-Why:
There's no -O0 hack there any more. Is this still a problem?


State-Changed-From-To: feedback->closed
State-Changed-By: jmcneill@NetBSD.org
State-Changed-When: Sat, 06 Jul 2019 17:01:27 +0000
State-Changed-Why:
Works with 8.99.50.


>Unformatted:

Home	PR Database Search