[Driver] Allow -e entry but reject -eentry

[MaskRay]

This short option taking an argument is unfortunate.

• If a cc1-only option starts with -e, using it for driver will not be
  reported as an error (e.g. commit
  6cd9886c88d16d288c74846495d89f2fe84ff827).
• If another -e driver option is intended but a typo is made, the
  option will be recognized as a -e.

gcc -export-dynamic passes -export-dynamic to ld. It's not clear
whether some options behave this way.

It seems -Wl,-eentry and -Wl,--entry=entry are primarily used. There
may also be a few gcc -e entry, but gcc -eentry is extremely rare or
not used at all. Therefore, we probably should reject the Joined form of
-e.

[llvmbot]

@llvm/pr-subscribers-clang-driver

@llvm/pr-subscribers-clang

Author: Fangrui Song (MaskRay)

Changes

This short option taking an argument is unfortunate.

• If a cc1-only option starts with -e, using it for driver will not be
  reported as an error (e.g. commit
  6cd9886c88d16d288c74846495d89f2fe84ff827).
• If another -e driver option is intended but a typo is made, the
  option will be recognized as a -e.

gcc -export-dynamic passes -export-dynamic to ld. It's not clear
whether some options behave this way.

It seems -Wl,-eentry and -Wl,--entry=entry are primarily used. There
may also be a few gcc -e entry, but gcc -eentry is extremely rare or
not used at all. Therefore, we probably should reject the Joined form of
-e.

---

Full diff: https://github.com/llvm/llvm-project/pull/72804.diff

2 Files Affected:

• (modified) clang/include/clang/Driver/Options.td (+1-1)
• (added) clang/test/Driver/entry.s (+5)

diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index df12ba8fbcb296a..83b09a892049bb6 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -1452,7 +1452,7 @@ def extract_api_ignores_EQ: CommaJoined<["--"], "extract-api-ignores=">,
   Visibility<[ClangOption, CC1Option]>,
     HelpText<"Comma separated list of files containing a new line separated list of API symbols to ignore when extracting API information.">,
     MarshallingInfoStringVector<FrontendOpts<"ExtractAPIIgnoresFileList">>;
-def e : JoinedOrSeparate<["-"], "e">, Flags<[LinkerInput]>, Group<Link_Group>;
+def e : Separate<["-"], "e">, Flags<[LinkerInput]>, Group<Link_Group>;
 def fmax_tokens_EQ : Joined<["-"], "fmax-tokens=">, Group<f_Group>,
   Visibility<[ClangOption, CC1Option]>,
   HelpText<"Max total number of preprocessed tokens for -Wmax-tokens.">,
diff --git a/clang/test/Driver/entry.s b/clang/test/Driver/entry.s
new file mode 100644
index 000000000000000..60ab89704c35462
--- /dev/null
+++ b/clang/test/Driver/entry.s
@@ -0,0 +1,5 @@
+/// To prevent mistaking -exxx as --entry=xxx, we allow -e xxx but reject -exxx.
+/// GCC -export-dynamic is rejected as well.
+// RUN: not %clang -### --target=x86_64-linux-gnu -export-dynamic %s 2>&1 | FileCheck %s
+
+// CHECK: error: unknown argument: '-export-dynamic'

[MaskRay]

Driver used to accept -e but ignore it (at least for Gnu.cpp). My fcf8ada on 2020-07-30 rendered -e (both Joined and Separate). I think if we impose a restriction that -exxx is unsupported, technically it is a breaking change but likely very few projects will notice it (and the issue is pretty straightforward to fix).

This change will get us out of the business to support random -exxx link option that GCC driver happens to accept but we don't, like -export-dynamic #72781 .

[dongjianqiang2]

LGTM.