Differences in building with gcc 3.4.6 versus gcc 4.4.0 ? - page 4 - General Discussion and Assistance - CHDK Forum

Differences in building with gcc 3.4.6 versus gcc 4.4.0 ?

  • 33 Replies
  • 24352 Views
Re: Differences in building with gcc 3.4.6 versus gcc 4.4.0 ?
« Reply #30 on: 26 / April / 2011, 13:14:29 »
Advertisements
Based on sub_FF95B0E4 in ixus120 103c, I'd guess dirent should be 64 bytes

edit
probably applies to all dryos R39 and later cams.

Can you tell for which instructions i need search ?
I test 64 bytes some times ago too, but on Ixus 1000 it not work.so i go back to 80
Ixus 1000 HS

*

Offline reyalp

  • ******
  • 14079
Re: Differences in building with gcc 3.4.6 versus gcc 4.4.0 ?
« Reply #31 on: 26 / April / 2011, 16:53:18 »
Based on sub_FF95B0E4 in ixus120 103c, I'd guess dirent should be 64 bytes

edit
probably applies to all dryos R39 and later cams.

Can you tell for which instructions i need search ?
I test 64 bytes some times ago too, but on Ixus 1000 it not work.so i go back to 80
Find a call to ReadFastDir in the Canon firmware, and check how much space is allocated for the second argument. It's usually on the stack, so the space is allocated with SUB SP, SP, #... which may include other things.

Looking at  sub_FF9A5570 ixus1000 100d, this is 0x40, like other >= 39 cams. It's possible you have other bugs ;)

I'm guessing that sub_FFAB9128 is ReadFastDir

Don't forget what the H stands for.

Re: Differences in building with gcc 3.4.6 versus gcc 4.4.0 ?
« Reply #32 on: 27 / April / 2011, 11:04:24 »
Based on sub_FF95B0E4 in ixus120 103c, I'd guess dirent should be 64 bytes

edit
probably applies to all dryos R39 and later cams.

Can you tell for which instructions i need search ?
I test 64 bytes some times ago too, but on Ixus 1000 it not work.so i go back to 80
Find a call to ReadFastDir in the Canon firmware, and check how much space is allocated for the second argument. It's usually on the stack, so the space is allocated with SUB SP, SP, #... which may include other things.

Looking at  sub_FF9A5570 ixus1000 100d, this is 0x40, like other >= 39 cams. It's possible you have other bugs ;)


I get the crash with 64 only when i use ptpcam GUI 0.2.66, and choose download all files or new files.
But the filebrowser or script select work with 64.

I dont know whats the current state of new ptpcam and GUI.when i update all files in source and use new ptpcam and GUI, is it possible that all can work for non expert as me and have same features as ptpcam GUI 0.2.66.Or is some important miss ?
Ixus 1000 HS

*

Offline reyalp

  • ******
  • 14079
Re: Differences in building with gcc 3.4.6 versus gcc 4.4.0 ?
« Reply #33 on: 27 / April / 2011, 12:36:35 »
Based on sub_FF95B0E4 in ixus120 103c, I'd guess dirent should be 64 bytes

edit
probably applies to all dryos R39 and later cams.

Can you tell for which instructions i need search ?
I test 64 bytes some times ago too, but on Ixus 1000 it not work.so i go back to 80
Find a call to ReadFastDir in the Canon firmware, and check how much space is allocated for the second argument. It's usually on the stack, so the space is allocated with SUB SP, SP, #... which may include other things.

Looking at  sub_FF9A5570 ixus1000 100d, this is 0x40, like other >= 39 cams. It's possible you have other bugs ;)


I get the crash with 64 only when i use ptpcam GUI 0.2.66, and choose download all files or new files.
But the filebrowser or script select work with 64.

I dont know whats the current state of new ptpcam and GUI.when i update all files in source and use new ptpcam and GUI, is it possible that all can work for non expert as me and have same features as ptpcam GUI 0.2.66.Or is some important miss ?
As far as I can tell (I may be wrong) the canon firmware only allocates 64 bytes for this structure, suggesting that your problem is elsewhere. It would be helpful to isolate what part of the ptpcam gui script crashes. A romlog might tell (should be accompanied by core/main.dump or core/main.bin from the same build)

As we have seen in this thread, moving things around it memory can hide/show other unrelated bugs.

If you update to 1152 the entire struct is moved into malloc'd memory, which may change things again ;)
Don't forget what the H stands for.


 

Related Topics