Hi,

OpenJDK project uses Cygwin to drive the OpenJDK build system under Windows. Recently, our GitHub 
Actions (GHA) Windows runs started to fail with make SEGV-ing:
     https://bugs.openjdk.java.net/browse/JDK-8276854

This also reproduces locally for me, with the standard OpenJDK Windows build. All sightings we had 
are on Windows 10, but it is unclear if it is Windows 10 specific.

The reproducer would be a bit hairy, because it requires MSVC 2019, JDK 17 binary, and some other 
dependencies to be installed for OpenJDK build. It goes like this:

------- 8< ------- ------- ------- ------- ------- ------- ------- -------

<open Cygwin shell>
$ git clone https://github.com/openjdk/jdk/ jdk
$ cd jdk
$ sh ./configure --with-boot-jdk=<jdk17-binary> --disable-warnings-as-errors
$ make clean jdk
    ...
Compiling 8 files for BUILD_TOOLS_LANGTOOLS
Compiling 1 files for BUILD_TOOLS_HOTSPOT
Compiling 12 properties into resource bundles for jdk.jdeps
/usr/bin/bash: line 1:  1588 Segmentation fault      (core dumped) /usr/bin/make -s -r -R -I
/cygdrive/d/a/jdk17u/jdk17u/jdk/make/common
SPEC=/cygdrive/d/a/jdk17u/jdk17u/jdk/build/windows-x64/spec.gmk MAKE_LOG_FLAGS="-s" -f
ModuleWrapper.gmk -I /cygdrive/d/a/jdk17u/jdk17u/jdk/make/common/modules
-I/cygdrive/d/a/jdk17u/jdk17u/jdk/make/modules/java.base MODULE=java.base MAKEFILE_PREFIX=Copy
make[2]: *** [make/Main.gmk:157: java.base-copy] Error 139
make[2]: *** Waiting for unfinished jobs....
Compiling 7 properties into resource bundles for jdk.jshell

------- 8< ------- ------- ------- ------- ------- ------- ------- -------

I have managed to bisect this in Cygwin to the difference between 3.2.0 and 3.3.0. Downgrading 
cygwin package to 3.2.0 workarounds this bug. This is the fix we are currently deploying in our 
build infra:
    https://github.com/openjdk/jdk/commit/403f3185f0988dcf6ef4e857d6659533bfa2943f

...but, of course, this is living on borrowed time, while 3.2.0 version is still available.

Any help would be appreciated. I have a working Windows VM where this issue reliably reproduces. 
Perhaps an easier way to follow up on this is to use me as remote hands.

After installing make-devel and doing "ulimit -c unlimited", I have got this stackdump:

Exception: STATUS_ACCESS_VIOLATION at rip=0018017516C
rax=0000000000000001 rbx=00000008004BE630 rcx=0000000000000001
rdx=00000000FFFFFFFF rsi=0000000800000000 rdi=0000000000000001
r8 =FFFFFFFFFFFFFFF1 r9 =0000000000000001 r10=0000000180320880
r11=0000000800000000 r12=0000000000000001 r13=0000000180243A80
r14=0000000000000000 r15=0000000000000000
rbp=0000000800403440 rsp=00000000FFFF99B0
program=C:\cygwin64\bin\make.exe, pid 61490, thread main
cs=0033 ds=002B es=002B fs=0053 gs=002B ss=002B
Stack trace:
Frame        Function    Args
00800403440  0018017516C (00100438940, 00800000000, 00000000001, 00800403440)
00800403440  001800D7C68 (000FFFF9BE7, 00000000000, 000FFFF9BE7, 008004035E0)
00800403440  0018018EB0B (000FFFF9BE7, 00000000000, 000FFFF9BE7, 008004035E0)
00800403440  0010040F513 (000FFFFE410, 001802A4CB0, 00180366E08, 00800071DB0)
00000000000  0010040F68C (00000000000, 001801402B0, 000FFFF9B20, 00100427E70)
00100427D70  001004104E3 (00000000000, 00000000000, 00000000006, 00000000000)
000FFFF9CD0  00100410D5F (7A01C9E544B96FAD, 00100419ED2, 001803231A0, 00800323460)
0080034D8A0  0010041C15F (00000000006, 00000000002, 00000000000, 00000000000)
00000000000  0010041C55E (00000000000, 00000000000, 00000000000, 00000000000)
00000000000  0010041B5B0 (00000000004, 006FFFFFFB9, 303E90000F032, 00180267740)
00000000000  0010041C55E (00000000000, 68F5F1EE9D7B, 00000000000, 00000000103)
00000000000  0010041B5B0 (00000000002, 00000000000, 00100000000, 00000000000)
00000000000  0010041C55E (000800D7C68, 00000000000, 000FFFFA2E8, 008001DA120)
00000000000  0010041B5B0 (53454400000000, 6176652D2A2D2824, 2D7367616C662D6C, 00000292D2A)
00100435590  0010041C985 (00000000000, 001803231A0, 00800075740, 00100429B21)
000FFFFB360  00100426324 (001801B609A, 00000000000, 00000000000, 000FFFFCD30)
000FFFFCD30  00180049B8D (00000000000, 00000000000, 00000000000, 00000000000)
000FFFFFFF0  00180047746 (00000000000, 00000000000, 00000000000, 00000000000)
000FFFFFFF0  001800477F4 (00000000000, 00000000000, 00000000000, 00000000000)
End of stack trace

It does not look particularly useful to me, I was hoping for symbol names to be resolved, to be 
honest. I don't know what I am supposed to do next. There is no "core" around, as far as I can see...

-- 
Thanks,
-Aleksey