Add option to disable BSS pad symbols

[andresj-sanchez]

Summary

• Add a --create-bss-pads / --no-create-bss-pads option to control autogenerated BSS/SBSS pad symbols.
• Keep the default enabled to preserve existing output.
• Avoid BSS range-check warnings when pad symbols are disabled and a declared symbol intentionally absorbs following padding.

Thanks to @dreamingmoths and @Mc-muffin for the help.

Context

Spimdisasm already has options for controlling autogenerated data and rodata pad symbols:

--create-data-pads / --no-create-data-pads
--create-rodata-pads / --no-create-rodata-pads

BSS currently always creates pad symbols after user-sized symbols. For example, a BSS symbol with declared size 0x4 followed by another symbol at +0x8 emits an autogenerated pad label for the gap:

dlabel some_bss_symbol
 .space 0x04

/* Automatically generated and unreferenced pad */
dlabel D_XXXXXXXX
 .space 0x04

This creates extra generated labels in BSS/SBSS output and can make objdiff views noisier.

Before:

After:

When those gap labels are not useful, disabling BSS pad creation folds the gap into the preceding BSS symbol instead:

dlabel some_bss_symbol
 .space 0x08

This PR adds the same style of control for BSS/SBSS via CREATE_BSS_PADS, exposed through the CLI as:

--create-bss-pads
--no-create-bss-pads

When enabled, behavior stays unchanged. When disabled, Spimdisasm stops creating those autogenerated BSS/SBSS pad symbols.

Notes

• This should be merged before the companion Splat PR that exposes create_bss_pads.
• Companion Splat PR: Expose spimdisasm pad creation options ethteck/splat#540

Testing

• Verified --no-create-bss-pads disables autogenerated BSS pad symbols.
• Verified --create-bss-pads preserves the existing autogenerated pad behavior.
• Verified zero-sized declared BSS symbols do not create pad symbols, matching data/rodata pad behavior.

[andresj-sanchez]

Wait @Mc-muffin found a possible bug with this approach related to the padding folding. I'll take a look at it.

[andresj-sanchez]

Fix pushed. @Mc-muffin was right: dropping the pad label also dropped the size boundary, so objdiff showed the previous BSS symbol absorbing the padding.

Now it follows the data/rodata-style end marker approach: emit the declared BSS size, close it with enddlabel, then leave the padding as unlabeled .space.

[AngheloAlf]

Funnily I was a bit worried something like this could happen, but you figured out on your own. I was planning on fixing it after the merge.

Either way, how does the generated assembly looks like? Hopefully you could show the before and after.

[andresj-sanchez]

Sure.

Before, with autogenerated pad label:

dlabel "m_tpcSingleton__27clsSingleton<11clsPfSystem>", weak
    /* 006E69B8 */ .space 0x04

/* Automatically generated and unreferenced pad */
dlabel D_006E69BC
    /* 006E69BC */ .space 0x04

Bad intermediate version:

dlabel "m_tpcSingleton__27clsSingleton<11clsPfSystem>", weak
    /* 006E69B8 */ .space 0x08

Current version:

dlabel "m_tpcSingleton__27clsSingleton<11clsPfSystem>", weak
    /* 006E69B8 */ .space 0x04
enddlabel "m_tpcSingleton__27clsSingleton<11clsPfSystem>"
    /* 006E69BC */ .space 0x04

So the dummy D_006E69BC label is gone, but the real BSS symbol still keeps its declared 0x04 size.

[AngheloAlf]

Awesome, that's exactly what I had in mind
thank you very much

[andresj-sanchez]

Thanks to you! and @Mc-muffin lol

[AngheloAlf]

fyi I can't make the new release today but I'll merge the pr either way.

I'll probably make the release tomorrow morning

[AngheloAlf]

hey @andresj-sanchez i got some time and made the release