[PATCH] D111101: [lld][WebAssembly] Remove redundant check for undefined global (NFC)

Heejin Ahn via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Oct 5 00:04:55 PDT 2021


aheejin added inline comments.


================
Comment at: lld/wasm/Relocations.cpp:45
-    if (g->importName)
-      return true;
   return config->allowUndefinedSymbols.count(sym->getName()) != 0;
----------------
sbc100 wrote:
> aheejin wrote:
> > sbc100 wrote:
> > > aheejin wrote:
> > > > sbc100 wrote:
> > > > > Actually there was a recent refactor that made `importName` a (optional) member of the Symbol superclass.. so I think this can be written as:
> > > > > 
> > > > > ```
> > > > > if (sym->importName)
> > > > >   return true;
> > > > > if (isa<UndefinedFunction>(sym) && config->importUndefined)
> > > > >   return true;
> > > > > ```
> > > > > 
> > > > > But if you would rather land this as is that is fine too.
> > > > > if (isa<UndefinedFunction>(sym) && config->importUndefined)
> > > > 
> > > > Isn't this `||` in the current code?
> > > > 
> > > > Also, if we return `true` unconditionally when `sym->importName` is defined, I think we return `true` for data or tables. Is that what you intended?
> > > Yes, any symbol with an explicit `importName` should always be imported and never have "undefined symbol" reported about it.
> > Which one is correct?
> > ```
> > if (isa<UndefinedFunction>(sym) && config->importUndefined)
> > ```
> > vs.
> > ```
> > if (isa<UndefinedFunction>(sym) || config->importUndefined)
> > ```
> > (The current code is like the latter and you suggested the former)
> > 
> > Also, `config->importUndefined` doesn't affect `UndefinedGlobal`, or `UndefinedTag` I am going to add?
> I think we should leave `importUndefined` as only effecting functions symbols for now.  We could followup by extending it but that would not be NFC.
> 
> Thus I think the `if (isa<UndefinedFunction>(sym) && config->importUndefined)` matches the original semantics and is that we should do here for now.
> 
1. You suggested to add
```
if (sym->importName)
  return true;
```
But doesn't this affect not only functions but all symbols?

2. 
> I think we should leave `importUndefined` as only effecting functions symbols for now.  We could followup by extending it but that would not be NFC.

The current code already handles globals, even twice, one of which I removed. So if we make this only handle functions that will not be NFC instead.. No?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D111101/new/

https://reviews.llvm.org/D111101



More information about the llvm-commits mailing list