FilterManager calls non-X SWI, and crashes without WindowManager

Hiya,

Simple one here…

On RISC OS 5.27, FilterManager calls a non-X SWI during its initialisation which will causes a bad crash when the SWI is not present. The SWI is Wimp_ReadSysInfo. This raises a warning report in my system (because non-X SWIs should not be called in privileged modes, except maybe in special circumstances), and then an error report because the environment handler is being called for an error from such a SWI call (which indicates that it’s not just a theoretical case, it actually happened).

Trace from the initialisation:

Initialise module: # 72 FilterManager            flags &00000001
Module instantiate: <Module(FilterManager @ &fc2b706c + &1398)>
Instantiate module FilterManager: class=<class 'riscos.modules.Instance'>
Calling module init for <Instance(FilterManager%Base, pw=&00000000)>
Init dispatch for module FilterManager%Base (instance 0) to &fc2b7564 (arguments='')
fc2b7564: {DA 'Extension ROM', module 'FilterManager': Entry Initialise}
fc2b7574: {DA 'Extension ROM', module 'FilterManager'}
==== Begin SWI misuse ====
SWI called without X-bit from non-USR mode: &400f2 (Wimp_ReadSysInfo)
Registers:
  r0  = &00000007, r1  = &00000000, r2  = &0700d4a4, r3  = &0000005c
  r4  = &00000000, r5  = &00000000, r6  = &00000000, r7  = &00000000
  r8  = &00000000, r9  = &00000000, r10 = &07007dcc, r11 = &00000000
  r12 = &0700d4a4, sp  = &04107f98, lr  = &04107fac, pc  = &fc2b75bc
  CPSR= &60000013 : SVC-32 ARM fi ae qvCZn
  SPSR= &40000013 : SVC-32 ARM fi ae qvcZn
Locations:
  r2  -> [&00000000, &00000000, &00000000, &00000000]
  r10 -> [&00000000, &00000000, &00000000, &00000000]
  pc is DA 'Extension ROM', module 'FilterManager'
  lr is DA 'SVC Stack'
Recently executed code:
  ---- Block &fc2b7564, 4 instructions ----
  fc2b7564: {DA 'Extension ROM', module 'FilterManager': Entry Initialise}
  fc2b7564: PUSH    {r0, r1, r2, r3, lr}
  fc2b7568: LDR     r2, [r12]
  fc2b756c: CMP     r2, #0
  fc2b7570: BNE     &FC2B758C
  ---- Block &fc2b7574, 3 instructions ----
  fc2b7574: {DA 'Extension ROM', module 'FilterManager'}
  fc2b7574: MOV     r0, #6
  fc2b7578: MOV     r3, #&5c                  ; #92  = '\'
  fc2b757c: SWI     XOS_Module
  ---- Block &fc2b7580, 2 instructions ----
  fc2b7580: ADDVS   sp, sp, #4
  fc2b7584: POPVS   {r1, r2, r3, pc}
  ---- Block &fc2b7588, 13 instructions ----
  fc2b7588: STR     r2, [r12]
  fc2b758c: MOV     r12, r2
  fc2b7590: MOV     r0, #0
  fc2b7594: STR     r0, [r12, #4]
  fc2b7598: STR     r0, [r12, #8]
  fc2b759c: STR     r0, [r12, #&c]
  fc2b75a0: STR     r0, [r12, #&10]
  fc2b75a4: STR     r0, [r12, #&14]
  fc2b75a8: STR     r0, [r12, #&18]
  fc2b75ac: STR     r0, [r12, #&40]
  fc2b75b0: STR     r0, [r12, #&54]
  fc2b75b4: MOV     r0, #7
  fc2b75b8: SWI     Wimp_ReadSysInfo
==== End SWI misuse ====
==== Begin error report ====
Error about to be reported to environment handler: [&1e6] SWI &400f2 (Wimp_ReadSysInfo) not known
  r0  = &00000007, r1  = &00000000, r2  = &0700d4a4, r3  = &0000005c
  r4  = &00000000, r5  = &00000000, r6  = &00000000, r7  = &00000000
  r8  = &00000000, r9  = &00000000, r10 = &07007dcc, r11 = &00000000
  r12 = &0700d4a4, sp  = &04107f98, lr  = &04107fac, pc  = &fc2b75bc
  CPSR= &60000013 : SVC-32 ARM fi ae qvCZn
  SPSR= &40000013 : SVC-32 ARM fi ae qvcZn
Locations:
  r2  -> [&00000000, &00000000, &00000000, &00000000]
  r10 -> [&00000000, &00000000, &00000000, &00000000]
  pc is DA 'Extension ROM', module 'FilterManager'
  lr is DA 'SVC Stack'
Recently executed code:
  ---- Block &fc2b7564, 4 instructions ----
  fc2b7564: {DA 'Extension ROM', module 'FilterManager': Entry Initialise}
  fc2b7564: PUSH    {r0, r1, r2, r3, lr}
  fc2b7568: LDR     r2, [r12]
  fc2b756c: CMP     r2, #0
  fc2b7570: BNE     &FC2B758C
  ---- Block &fc2b7574, 3 instructions ----
  fc2b7574: {DA 'Extension ROM', module 'FilterManager'}
  fc2b7574: MOV     r0, #6
  fc2b7578: MOV     r3, #&5c                  ; #92  = '\'
  fc2b757c: SWI     XOS_Module
  ---- Block &fc2b7580, 2 instructions ----
  fc2b7580: ADDVS   sp, sp, #4
  fc2b7584: POPVS   {r1, r2, r3, pc}
  ---- Block &fc2b7588, 13 instructions ----
  fc2b7588: STR     r2, [r12]
  fc2b758c: MOV     r12, r2
  fc2b7590: MOV     r0, #0
  fc2b7594: STR     r0, [r12, #4]
  fc2b7598: STR     r0, [r12, #8]
  fc2b759c: STR     r0, [r12, #&c]
  fc2b75a0: STR     r0, [r12, #&10]
  fc2b75a4: STR     r0, [r12, #&14]
  fc2b75a8: STR     r0, [r12, #&18]
  fc2b75ac: STR     r0, [r12, #&40]
  fc2b75b0: STR     r0, [r12, #&54]
  fc2b75b4: MOV     r0, #7
  fc2b75b8: SWI     Wimp_ReadSysInfo
==== End error report ====
Instantiate failed: RISCOSSyntheticError('[&1e6] SWI &400f2 (Wimp_ReadSysInfo) not known',)
  Error: [&1e6] SWI &400f2 (Wimp_ReadSysInfo) not known

Report from FilterManager 0.29; entry from *ROMModules:

Extension ROM 1 :     73 : fc2b706c : Dormant  : 0.29     : FilterManager

On RISC OS 5.27, FilterManager calls a non-X SWI during its initialisation which will causes a bad crash when the SWI is not present. The SWI is Wimp_ReadSysInfo.

We might assume that has carried through to the current dev stream (v5.29) and is “baked in” on v5.28 stable, but maybe not.

v5.27 is essentially history. Useful to step back to if a (apparently) new bug appears, otherwise – history.

It can be a bit of a pain keeping up with the changes at times, not too bad in recent weeks though.

I’m sorry, I’m not sure what the point of your reply was.

Are you saying that the report is not worthwhile because it’s about a ‘historic’ version?

While it is good practice to call X form SWIs when initialising ¹, a part of me thinks that this “bug” is mostly happening due to a peculiarity of your setup. Namely, that your system appears to be trying to initialise a module that interacts with the Wimp environment on a system that doesn’t have the Wimp running, which would be “unusual”. ²

Still worth adding to the bug tracker though, because wrong is wrong.

¹ Good form, as in the documentation (PRM p1-213) basically says do not generate errors and always call X form SWIs. ;-)

² Makes me wonder if there are similar situations in other parts of the Desktop world, like Switcher or…..

I’m sorry, I’m not sure what the point of your reply was.
Are you saying that the report is not worthwhile because it’s about a ‘historic’ version?

Ooooh, no, no, no.
Just concerned that you might end up chasing something that had already been fixed.

I think it’s clearer if you read along with my comment on another post about stating the version, including a specific release date, that you are testing on.

There are few enough people with time, inclination and talent doing anything for RO, so I wouldn’t want anyone to be chasing down something that was already found and fixed.

Example items in the cut & paste work done in the latter months of 5.27 releases where different aspects might be faulty one day and fixed the next and possibly a fix a day over a week.

A quick look confirms it… https://gitlab.riscosopen.org/RiscOS/Sources/Desktop/Filter/-/blob/master/s/ModHead#L500

PS: I can’t add it to the bug tracker. It keeps asking me to log in, reporting that my login is unsuccessful (while the login to everything else works). Hmm… 🤔

This bug remains in RISC OS 5.31 RC3, and presents in exactly the same way…

Initialise module: # 72 FilterManager            flags &00000001
==== Begin SWI misuse ====
SWI called without X-bit from non-USR mode: &400f2 (Wimp_ReadSysInfo)
Registers:
  r0  = &00000007, r1  = &00000000, r2  = &07019d9c, r3  = &0000005c
  r4  = &00000000, r5  = &00000000, r6  = &00000000, r7  = &00000000
  r8  = &00000000, r9  = &00000000, r10 = &07012c9c, r11 = &00000000
  r12 = &07019d9c, sp  = &04107f98, lr  = &04107fac, pc  = &fc2ba24c
  CPSR= &60000013 : SVC-32 ARM fi ae qvCZn
  SPSR= &40000013 : SVC-32 ARM fi ae qvcZn
Locations:
  r2  -> [&00000000, &00000000, &00000000, &00000000] in DA 'Module area', module 'FilterManager%Base' workspace
  r10 -> [&00000000, &000000d9, &fc10bfa0, &00018c48] in DA 'Module area'
  r12 -> [&00000000, &00000000, &00000000, &00000000] in DA 'Module area', module 'FilterManager%Base' workspace
  pc is DA 'Extension ROM', module 'FilterManager'
  lr is DA 'SVC Stack'
Recently executed code:
  ---- Block &fc2ba1f4, 4 instructions ----
  fc2ba1f4: {DA 'Extension ROM', module 'FilterManager': Entry Initialise}
  fc2ba1f4: PUSH    {r0, r1, r2, r3, lr}
  fc2ba1f8: LDR     r2, [r12]
  fc2ba1fc: CMP     r2, #0
  fc2ba200: BNE     &FC2BA21C
  ---- Block &fc2ba204, 3 instructions ----
  fc2ba204: {DA 'Extension ROM', module 'FilterManager'}
  fc2ba204: MOV     r0, #6
  fc2ba208: MOV     r3, #&5c                  ; #92  = '\'
  fc2ba20c: SWI     XOS_Module
  ---- Block &fc2ba210, 2 instructions ----
  fc2ba210: ADDVS   sp, sp, #4
  fc2ba214: POPVS   {r1, r2, r3, pc}
  ---- Block &fc2ba218, 13 instructions ----
  fc2ba218: STR     r2, [r12]
  fc2ba21c: MOV     r12, r2
  fc2ba220: MOV     r0, #0
  fc2ba224: STR     r0, [r12, #4]
  fc2ba228: STR     r0, [r12, #8]
  fc2ba22c: STR     r0, [r12, #&c]
  fc2ba230: STR     r0, [r12, #&10]
  fc2ba234: STR     r0, [r12, #&14]
  fc2ba238: STR     r0, [r12, #&18]
  fc2ba23c: STR     r0, [r12, #&40]
  fc2ba240: STR     r0, [r12, #&54]
  fc2ba244: MOV     r0, #7
  fc2ba248: SWI     Wimp_ReadSysInfo
==== End SWI misuse ====

Regarding Rick’s comment about it being a pecularity of my system…

Not having a WindowManager (or not having ANY module) might be odd, but as you also say, it’s wrong to have such SWIs present. Failng to handle such cases it causes obscure problems during reinitialisation, or if softload is happening, or if someone is just trying to use a system that isn’t in the same configuration as the default distribution.