[PATCH v2] ld: Rename a file on Windows fails if target already exists

[Jan Beulich]

On 19.07.2025 09:34, Torbjörn SVENSSON wrote:
> Changes since v1:
> 
> - Apparently, the include of errno was never committed before.
>   I've rebuilt and rechecked that this works on Windows again.
> 
> 
> Without this patch, the GCC testsuite gives errors like:
> 
> Testing gcc.dg/lto/20091027-1, -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin
> doing compile
> Executing on host: arm-none-eabi-gcc.exe   -mthumb -march=armv6s-m -mcpu=cortex-m0 -mfloat-abi=soft   -fdiagnostics-plain-output  -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin   -c     -o c_lto_20091027-1_0.o /build/src/gcc/gcc/testsuite/gcc.dg/lto/20091027-1_0.c    (timeout = 800)
> spawn -ignore SIGHUP arm-none-eabi-gcc.exe -mthumb -march=armv6s-m -mcpu=cortex-m0 -mfloat-abi=soft -fdiagnostics-plain-output -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin -c -o c_lto_20091027-1_0.o /build/src/gcc/gcc/testsuite/gcc.dg/lto/20091027-1_0.c
> pid is 52421 -52421
> pid is -1
> output is  status 0
> PASS: gcc.dg/lto/20091027-1 c_lto_20091027-1_0.o assemble, -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin
> doing compile
> Executing on host: arm-none-eabi-gcc.exe   -mthumb -march=armv6s-m -mcpu=cortex-m0 -mfloat-abi=soft   -fdiagnostics-plain-output  -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin   -c     -o c_lto_20091027-1_1.o /build/src/gcc/gcc/testsuite/gcc.dg/lto/20091027-1_1.c    (timeout = 800)
> spawn -ignore SIGHUP arm-none-eabi-gcc.exe -mthumb -march=armv6s-m -mcpu=cortex-m0 -mfloat-abi=soft -fdiagnostics-plain-output -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin -c -o c_lto_20091027-1_1.o /build/src/gcc/gcc/testsuite/gcc.dg/lto/20091027-1_1.c
> pid is 52444 -52444
> pid is -1
> output is  status 0
> PASS: gcc.dg/lto/20091027-1 c_lto_20091027-1_1.o assemble, -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin
> doing compile
> Executing on host: arm-none-eabi-gcc.exe  c_lto_20091027-1_0.o c_lto_20091027-1_1.o  -mthumb -march=armv6s-m -mcpu=cortex-m0 -mfloat-abi=soft   -dumpbase "" -fdiagnostics-plain-output  -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin  -r -nostdlib -flinker-output=nolto-rel      -Wl,--start-group -lc -lm -Wl,--end-group --specs=nosys.specs      -o gcc-dg-lto-20091027-1-21.exe    (timeout = 800)
> spawn -ignore SIGHUP arm-none-eabi-gcc.exe c_lto_20091027-1_0.o c_lto_20091027-1_1.o -mthumb -march=armv6s-m -mcpu=cortex-m0 -mfloat-abi=soft -dumpbase  -fdiagnostics-plain-output -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin -r -nostdlib -flinker-output=nolto-rel -Wl,--start-group -lc -lm -Wl,--end-group --specs=nosys.specs -o gcc-dg-lto-20091027-1-21.exe
> pid is 52471 -52471
> T:../lib/gcc/arm-none-eabi/14.3.1/../../../../arm-none-eabi/bin/ld.exe: failed to rename output with object-only section
> collect2.exe: error: ld returned 1 exit status
> pid is -1
> close result is 52471 exp6 0 1
> output is T:../lib/gcc/arm-none-eabi/14.3.1/../../../../arm-none-eabi/bin/ld.exe: failed to rename output with object-only section
> collect2.exe: error: ld returned 1 exit status
>  status 1
> compiler exited with status 1
> FAIL: gcc.dg/lto/20091027-1 c_lto_20091027-1_0.o-c_lto_20091027-1_1.o link, -O0 -flto -flto-partition=1to1 -fno-use-linker-plugin
> 
> 
> With this patch, the test is a PASS.
> 
> Ok for trunk?

Yes.

> @Nick: Is this something that should be considered for 2.45 branch? Regression started with 2.44 release.

Cc-ing Nick for him to answer this.

Jan