RISC OS Open: Forum: ZeroPain2

Nov 2, 2016 11:03am

Colin Ferris (399) 1818 posts

Is there a ZeroPain version of Sargasso?

Is there a standard way of updating ZeroPain problems – ie in a !RunImage – teq rx,#0 and bypassing the problem?

(working on the !RunImage a large Asm file)

Nov 2, 2016 8:20pm

Rick Murray (539) 13855 posts

Is there a standard way of updating ZeroPain problems

Your TEQ might work as a sanity check in a module or the like, however in my experience of tracking down page zero issues, it is usually an actual bug that needs sorting out – usually either references to a handle of array before it has been assigned, or after it has been freed… or screwy logic that tries to be clever and do things like “if something is NULL or something.somethingelse is NULL”…(the second clause running foul of trying to look at a potentially NULL pointer!).

As such, a program that lives in &8000-&xxxxx (or a module further away) accessing memory around &0 is a clear sign that something is going wrong. Better to cure the cause than to provide a workaround elsewhere.

Though, if you’re doing a patch-up job to a disassembled executable, your job is suddenly much harder, and in this case, one could overlook some TEQs in lieu of having any actual source to play with. :-)

Nov 9, 2016 9:29am

Colin Ferris (399) 1818 posts

Thanks for the reply – I presume if you have the source – the ‘C’ compiler won’t flag an error – if it produces code that peeks the zero page.

Are there any examples of the problem code available in ‘C’- and also the corrected versions?

Are there any maintainers of ‘’ v2.03 31 Aug 2013 and ’POPstar’ v2.06-ds.5 (8-Jun-2013) – both producing zero pain output.
**********
POPstar
R0 = 0001a2a8 R1 = 00000000 R2 = 00000020 R3 = 00000000
R4 = 0001a008 R5 = 2024bad9 R6 = 2024b919 R7 = 00000000
R8 = 00000000 R9 = 0001a2c8 R10 = 0001d1a0 R11 = 0001ded0
R12 = 0001de24 R13 = 0001de88 R14 = 0000af30 R15 = 00008c24
DFAR = 00000000 Mode USR32 Flags nZCv if PSR = 60000010

00008c14 : e7c0c003 : STRB R12,[R0,R3]
00008c18 : e2833001 : ADD R3,R3,#1
00008c1c * e7d1c003 * LDRB R12,[R1,R3]
00008c20 : e35c001f : CMP R12,#&1F
00008c24 : 9a000001 : BLS &00008C30
*******
Sargasso
R0 = 004d00e0 R1 = 004cffd0 R2 = 00000000 R3 = 004ce9c8
R4 = 0043d56c R5 = 00000032 R6 = 0043d0ec R7 = 00000004
R8 = 004d0080 R9 = 003e4048 R10 = 003f3208 R11 = 003f3c78
R12 = 004d0010 R13 = 003f3bfc R14 = 000000d0 R15 = 0000ab4c
DFAR = 00000000 Mode USR32 Flags nzCv if PSR = 20000010

0000ab38 : e5841048 : STR R1,[R4,#72]
0000ab3c : 0a000003 : BEQ &0000AB50
0000ab40 : e59d2028 : LDR R2,[R13,#40]
0000ab44 * e5d23000 * LDRB R3,[R2,#0]
0000ab48 : e353003c : CMP R3,#&3C ; =“<”
0000ab4c : 0a00000e : BEQ &0000AB8C

Nov 9, 2016 9:41am

John Williams (567) 768 posts

James Bursa (Sargasso) and Darren Salt (ds versions of POPstar).

James is part of the NetSurf team, and Darren, once of YouMustBeJoking, has moved sites but should be easily traceable.

But are you sure it’s really these apps causing the problem? I don’t actually understand the ZeroPain output myself!

Nov 9, 2016 7:22pm

Steve Pampling (1551) 8173 posts

Are there any maintainers of ‘’ v2.03 31 Aug 2013 and ’POPstar’ v2.06-ds.5 (8-Jun-2013)

Well, the source for PopStar is available, at the same location as the executable. (IIRC each person modifying the original has complied with the source availability requirement.

ZeroPain2

Reply

Search forums

Social

ROOL Store

Donate! Why?

RISC OS IPR

Description

Voices

Options

Nov 2, 2016 11:03am Colin Ferris (399) 1818 posts	Is there a ZeroPain version of Sargasso? Is there a standard way of updating ZeroPain problems – ie in a !RunImage – teq rx,#0 and bypassing the problem? (working on the !RunImage a large Asm file)

Nov 2, 2016 8:20pm Rick Murray (539) 13855 posts	Is there a standard way of updating ZeroPain problems Your TEQ might work as a sanity check in a module or the like, however in my experience of tracking down page zero issues, it is usually an actual bug that needs sorting out – usually either references to a handle of array before it has been assigned, or after it has been freed… or screwy logic that tries to be clever and do things like “if something is NULL or something.somethingelse is NULL”…(the second clause running foul of trying to look at a potentially NULL pointer!). As such, a program that lives in &8000-&xxxxx (or a module further away) accessing memory around &0 is a clear sign that something is going wrong. Better to cure the cause than to provide a workaround elsewhere. Though, if you’re doing a patch-up job to a disassembled executable, your job is suddenly much harder, and in this case, one could overlook some TEQs in lieu of having any actual source to play with. :-)

Nov 9, 2016 9:29am Colin Ferris (399) 1818 posts	Thanks for the reply – I presume if you have the source – the ‘C’ compiler won’t flag an error – if it produces code that peeks the zero page. Are there any examples of the problem code available in ‘C’- and also the corrected versions? Are there any maintainers of ‘’ v2.03 31 Aug 2013 and ’POPstar’ v2.06-ds.5 (8-Jun-2013) – both producing zero pain output. ********** POPstar R0 = 0001a2a8 R1 = 00000000 R2 = 00000020 R3 = 00000000 R4 = 0001a008 R5 = 2024bad9 R6 = 2024b919 R7 = 00000000 R8 = 00000000 R9 = 0001a2c8 R10 = 0001d1a0 R11 = 0001ded0 R12 = 0001de24 R13 = 0001de88 R14 = 0000af30 R15 = 00008c24 DFAR = 00000000 Mode USR32 Flags nZCv if PSR = 60000010 00008c14 : e7c0c003 : STRB R12,[R0,R3] 00008c18 : e2833001 : ADD R3,R3,#1 00008c1c * e7d1c003 * LDRB R12,[R1,R3] 00008c20 : e35c001f : CMP R12,#&1F 00008c24 : 9a000001 : BLS &00008C30 ******* Sargasso R0 = 004d00e0 R1 = 004cffd0 R2 = 00000000 R3 = 004ce9c8 R4 = 0043d56c R5 = 00000032 R6 = 0043d0ec R7 = 00000004 R8 = 004d0080 R9 = 003e4048 R10 = 003f3208 R11 = 003f3c78 R12 = 004d0010 R13 = 003f3bfc R14 = 000000d0 R15 = 0000ab4c DFAR = 00000000 Mode USR32 Flags nzCv if PSR = 20000010 0000ab38 : e5841048 : STR R1,[R4,#72] 0000ab3c : 0a000003 : BEQ &0000AB50 0000ab40 : e59d2028 : LDR R2,[R13,#40] 0000ab44 * e5d23000 * LDRB R3,[R2,#0] 0000ab48 : e353003c : CMP R3,#&3C ; =“<” 0000ab4c : 0a00000e : BEQ &0000AB8C

Nov 9, 2016 9:41am John Williams (567) 768 posts	James Bursa (Sargasso) and Darren Salt (ds versions of POPstar). James is part of the NetSurf team, and Darren, once of YouMustBeJoking, has moved sites but should be easily traceable. But are you sure it’s really these apps causing the problem? I don’t actually understand the ZeroPain output myself!

Nov 9, 2016 7:22pm Steve Pampling (1551) 8173 posts	Are there any maintainers of ‘’ v2.03 31 Aug 2013 and ’POPstar’ v2.06-ds.5 (8-Jun-2013) Well, the source for PopStar is available, at the same location as the executable. (IIRC each person modifying the original has complied with the source availability requirement.