On 2026-04-22 20:46, Mike Stump wrote:
On Apr 15, 2026, at 10:08 PM, Torbjorn SVENSSON <[email protected]>
wrote:
Gentle ping! :)
Ok.
Pushed as r17-407-g8e1592700071f0 and r16-8887-gc2ad32446c8e13.
Kind regards,
Torbjörn
Kind regards,
Torbjörn
On 2026-04-08 13:38, Torbjörn SVENSSON wrote:
Without this patch, I see errors like:
FAIL: g++.dg/modules/compile-std1.C module-cmi <bits/stdc++.h>
(gcm.cache/,/T:/build/r16-8473-g5cc0ead3625fe6/arm-none-eabi/include/c++/16.0.1/arm-none-eabi/thumb/v6-m/nofp/bits/stdc++.h.gcm)
Obviously, the path is illegal on Windows since the colon is a reserved
character.
With below patch, I instead get:
PASS: g++.dg/modules/compile-std1.C module-cmi <bits/stdc++.h>
(gcm.cache/T-/build/r16-8473-g5cc0ead3625fe6/arm-none-eabi/include/c++/16.0.1/arm-none-eabi/thumb/v6-m/nofp/bits/stdc++.h.gcm)
There is no impact of this change on Linux when testing an arm-none-eabi
toolchain.
Ok for trunk?
--
GCC munges absolute paths on Windows by replacing the colon with a dash.
gcc/testsuite/ChangeLog:
* g++.dg/modules/modules.exp: Replace colon with dash on
Windows.
Signed-off-by: Torbjörn SVENSSON <[email protected]>
---
gcc/testsuite/g++.dg/modules/modules.exp | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/gcc/testsuite/g++.dg/modules/modules.exp
b/gcc/testsuite/g++.dg/modules/modules.exp
index 2c7365db1cc..aa351b73e0e 100644
--- a/gcc/testsuite/g++.dg/modules/modules.exp
+++ b/gcc/testsuite/g++.dg/modules/modules.exp
@@ -83,7 +83,10 @@ proc host_header_path {hname} {
# Return the pathname CMI munged like the compiler.
proc munge_cmi {cmi} {
- if { [string index $cmi 0] == "/" } {
+ if { [string index $cmi 1] == ":" && [info exists ::env(OS)] && [string match
"Windows*" $::env(OS)] } {
+ # Windows paths like C:/some/path should be munged to C-/some/path
+ set cmi [string replace $cmi 1 1 "-"]
+ } elseif { [string index $cmi 0] == "/" } {
set cmi [string range $cmi 1 end]
} else {
set cmi ",/$cmi"
