[PATCH] D76250: [llvm-objdump] Only reject -long-option in objdump mode

[James Henderson via Phabricator via llvm-commits]

jhenderson added a comment.

In D76250#1958357 <https://reviews.llvm.org/D76250#1958357>, @compnerd wrote:

> The precedent for behaviour based on the name is long standing and not being introduced here.  In fact the archiver already does this.  `clang` also does this (`clang` vs `clang++`).  This is not at all uncommon I think.  The target triple prefixed tools should be treated as the tool that they specify and implicitly passing `-target ...` which is derived from the prefix.


I'm well aware that there's precedent for it, but I'm also aware that this precedent has caused problems, as already highlighted. It also leads to a plethora of tools in the build output, in order to test all the behaviours at runtime and (presumably) to simplify packaging/installation etc. These copies, whilst on some platforms (Windows) are not symlinks, leading to multiple copies of the same tool (with different names) that are several hundred MBs in size in the build output.

@compnerd - how do you propose resolving the ambiguity of multiple short options, grouped together versus a single-long option, if single-dash is allowed for long options too?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D76250/new/

https://reviews.llvm.org/D76250