Small IRA Tutorial

[phx]

Originally Posted by hop

it would be preferable not to discard known symbol data from HUNK_SYMBOL if possible. The config file symbols do not have an associated section, but HUNK_SYMBOLs do, would it be possible to parse and store this information to use in this situation?

Yes, of course you can. I even wrote the code today to do that - although I think it is not worth the effort, just for a symbol at the beginning of a section which doesn't show up when the previous section is empty. How often is that a problem?

In most cases executables have neither symbols nor empty sections. And then you still have the

SECSTRT_n

label.

At the end of the day I had to realise that I still cannot determine if a label should be generated for the empty section or the following section, because the offset is identical. Label-generation is triggered by offsets stored in the

ira->LabelAdr2[]

array. And its index (

p2labind

) will be incremented as long as the current offset matches.

Nothing is impossible, and I might have another look this week, but I fear the solution of this problem would need much bigger modifications.

[kamelito]

Dunno how often but as it has been generated by an old C compiler it could be seen elsewhere I guess.

[hop]

Originally Posted by phx

Yes, of course you can. I even wrote the code today to do that - although I think it is not worth the effort, just for a symbol at the beginning of a section which doesn't show up when the previous section is empty. How often is that a problem?

In most cases executables have neither symbols nor empty sections. And then you still have the

SECSTRT_n

label.

At the end of the day I had to realise that I still cannot determine if a label should be generated for the empty section or the following section, because the offset is identical. Label-generation is triggered by offsets stored in the

ira->LabelAdr2[]

array. And its index (

p2labind

) will be incremented as long as the current offset matches.

Nothing is impossible, and I might have another look this week, but I fear the solution of this problem would need much bigger modifications.

Thank you very much for taking the time to look into this. I appreciate that it is not a high-priority issue, but in the interest of correctness I think it is worth doing, if it is not too much work. As @kamelito says, this may be a feature of a common toolchain, although shipping with debug symbols in place may be less common!

Would it be possible to add another array alongside

ira->LabelAdr2[]

of equal size called something like

ira->LabelHunkIndex[]

? The elements would default to -1 (unknown hunk) but when symbols are added from a

HUNK_SYMBOL

it could be set to a valid value. Then this could be read in

WriteSection

[kamelito]

There is many games and some apps that contains symbols.

[phx]

Originally Posted by hop

in the interest of correctness I think it is worth doing, if it is not too much work.

Indeed, it would require some work. I'm still not sure if it is worth it.

Also I cannot promise I will have a look at it soon, because I just arrived at my holiday site, where I will spend the next 14 days. Even if bad weather may leave me some time (which is likely), my development equipment here is not optimal.

Would it be possible to add another array alongside

ira->LabelAdr2[]

of equal size called something like

ira->LabelHunkIndex[]

?

Yes, this might be possible, and I thought about it too. Although I don't like it very much. It is like "shooting sparrows with cannons", as we say in german.

Originally Posted by kamelito

There is many games and some apps that contains symbols.

Yes, and this works fine. But how often do you have such a symbol at the section's base address, following an empty section? There are probably more serious problems to tackle.

[hop]

Have a great holiday.

Originally Posted by phx

Indeed, it would require some work. I'm still not sure if it is worth it.

I agree it's an edge case.

I'm starting to gravitate towards supporting structured data however...

[kamelito]

Originally Posted by hop

Have a great holiday.



I agree it's an edge case.

I'm starting to gravitate towards supporting structured data however...

@hop
Are you going to use the include to define structs used for OS functions calls?