ARROW-4744: [C++][CI] Change mingw builds back to debug. Cleanup up some version warnings

[emkornfield]

No description provided.

[kou]

Thanks!
Should we do the same thing for Zstd?

https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22762420/job/4xuv4mpnwvmq5byf#L391

C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc: In member function 'virtual int64_t arrow::util::ZSTDCodec::MaxCompressedLen(int64_t, const uint8_t*)':
C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc:232:38: error: conversion to 'size_t {aka unsigned int}' from 'int64_t {aka long long int}' may alter its value [-Werror=conversion]
return ZSTD_compressBound(input_len);
^
cc1plus.exe: all warnings being treated as errors

[emkornfield]

Yes on it.

[emkornfield]

Also, should/can these checks be turned on for linux builds?

[emkornfield]

or turned off for ming in the short term? I don't understand why the breakage just started occurring now.

[kou]

It's caused by combination of #3693 and #3750. It's my fault. Sorry.
We can disable -Werror by the following change:

diff --git a/ci/appveyor-cpp-build-mingw.bat b/ci/appveyor-cpp-build-mingw.bat
index 249ea1ab..bdd3b144 100644
--- a/ci/appveyor-cpp-build-mingw.bat+++ b/ci/appveyor-cpp-build-mingw.bat@@ -17,8 +17,8 @@
@echo on
-set CMAKE_BUILD_TYPE=debug-set MESON_BUILD_TYPE=debug+set CMAKE_BUILD_TYPE=release+set MESON_BUILD_TYPE=release
set INSTALL_DIR=%HOMEDRIVE%%HOMEPATH%\install
set PATH=%INSTALL_DIR%\bin;%PATH%

[emkornfield]

@kou I made the change, lets hope that fixes CI (I think the other changes aren't bad so I'll leave them in if that is OK). I sent a follow-up e-mail to the mailing to see how we want to proceed with some of these ignored warnings.

[emkornfield]

Build failure is plasma (looks like it might be flaky? because its based on timestamps):

 # Test timestamps
t1 = time.time()
x, _, _ = create_object(plasma_client, 3, metadata_size=0, seal=False)
t2 = time.time()
l6 = plasma_client.list()

 assert math.floor(t1) <= l6[x]["create_time"] <= math.ceil(t2)

E assert 1551585976 <= 1551585975
E + where 1551585976 = (1551585976.0053494)
E + where = math.floor

[kou]

+1
arrow-array-test is stuck with MinGW 32bit but it's not related in this change.
https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779178/job/787pa80bsth5rb4d#L727

[xhochy]

@kou@emkornfield The 64bit builds now get stuck from time to time too on master: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779601/job/d1h0bfv5oyecs414

[emkornfield]

@kou do you have thoughts on what might be causing that? Could we maybe disable ming64 until we can figure it out?

[kou]

I don't know why... It's not reproduced on my local machine.
We can disable only arrow-array-test instead of disabling all MinGW 64bit build. I want to keep Apache Arrow 0.13.0 buildable with MinGW.

[emkornfield]

@kou is disabling something you can do? I'm currently working on another flaky test and probably won't be able to get to it tonight. (otherwise I can try to do it tomorrow)

[kou]

OK. I'll do later.

[kou]

Done: #3809

[emkornfield]

No description provided.

[kou]

Thanks!
Should we do the same thing for Zstd?

https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22762420/job/4xuv4mpnwvmq5byf#L391

C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc: In member function 'virtual int64_t arrow::util::ZSTDCodec::MaxCompressedLen(int64_t, const uint8_t*)':
C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc:232:38: error: conversion to 'size_t {aka unsigned int}' from 'int64_t {aka long long int}' may alter its value [-Werror=conversion]
return ZSTD_compressBound(input_len);
^
cc1plus.exe: all warnings being treated as errors

[emkornfield]

Yes on it.

[emkornfield]

Also, should/can these checks be turned on for linux builds?

[emkornfield]

or turned off for ming in the short term? I don't understand why the breakage just started occurring now.

[kou]

It's caused by combination of #3693 and #3750. It's my fault. Sorry.
We can disable -Werror by the following change:

diff --git a/ci/appveyor-cpp-build-mingw.bat b/ci/appveyor-cpp-build-mingw.bat
index 249ea1ab..bdd3b144 100644
--- a/ci/appveyor-cpp-build-mingw.bat+++ b/ci/appveyor-cpp-build-mingw.bat@@ -17,8 +17,8 @@
@echo on
-set CMAKE_BUILD_TYPE=debug-set MESON_BUILD_TYPE=debug+set CMAKE_BUILD_TYPE=release+set MESON_BUILD_TYPE=release
set INSTALL_DIR=%HOMEDRIVE%%HOMEPATH%\install
set PATH=%INSTALL_DIR%\bin;%PATH%

[emkornfield]

@kou I made the change, lets hope that fixes CI (I think the other changes aren't bad so I'll leave them in if that is OK). I sent a follow-up e-mail to the mailing to see how we want to proceed with some of these ignored warnings.

[emkornfield]

Build failure is plasma (looks like it might be flaky? because its based on timestamps):

 # Test timestamps
t1 = time.time()
x, _, _ = create_object(plasma_client, 3, metadata_size=0, seal=False)
t2 = time.time()
l6 = plasma_client.list()

 assert math.floor(t1) <= l6[x]["create_time"] <= math.ceil(t2)

E assert 1551585976 <= 1551585975
E + where 1551585976 = (1551585976.0053494)
E + where = math.floor

[kou]

+1
arrow-array-test is stuck with MinGW 32bit but it's not related in this change.
https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779178/job/787pa80bsth5rb4d#L727

[xhochy]

@kou@emkornfield The 64bit builds now get stuck from time to time too on master: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779601/job/d1h0bfv5oyecs414

[emkornfield]

@kou do you have thoughts on what might be causing that? Could we maybe disable ming64 until we can figure it out?

[kou]

I don't know why... It's not reproduced on my local machine.
We can disable only arrow-array-test instead of disabling all MinGW 64bit build. I want to keep Apache Arrow 0.13.0 buildable with MinGW.

[emkornfield]

@kou is disabling something you can do? I'm currently working on another flaky test and probably won't be able to get to it tonight. (otherwise I can try to do it tomorrow)

[kou]

OK. I'll do later.

[kou]

Done: #3809

[emkornfield]

No description provided.

[kou]

Thanks!
Should we do the same thing for Zstd?

https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22762420/job/4xuv4mpnwvmq5byf#L391

C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc: In member function 'virtual int64_t arrow::util::ZSTDCodec::MaxCompressedLen(int64_t, const uint8_t*)':
C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc:232:38: error: conversion to 'size_t {aka unsigned int}' from 'int64_t {aka long long int}' may alter its value [-Werror=conversion]
return ZSTD_compressBound(input_len);
^
cc1plus.exe: all warnings being treated as errors

[emkornfield]

Yes on it.

[emkornfield]

Also, should/can these checks be turned on for linux builds?

[emkornfield]

or turned off for ming in the short term? I don't understand why the breakage just started occurring now.

[kou]

It's caused by combination of #3693 and #3750. It's my fault. Sorry.
We can disable -Werror by the following change:

diff --git a/ci/appveyor-cpp-build-mingw.bat b/ci/appveyor-cpp-build-mingw.bat
index 249ea1ab..bdd3b144 100644
--- a/ci/appveyor-cpp-build-mingw.bat+++ b/ci/appveyor-cpp-build-mingw.bat@@ -17,8 +17,8 @@
@echo on
-set CMAKE_BUILD_TYPE=debug-set MESON_BUILD_TYPE=debug+set CMAKE_BUILD_TYPE=release+set MESON_BUILD_TYPE=release
set INSTALL_DIR=%HOMEDRIVE%%HOMEPATH%\install
set PATH=%INSTALL_DIR%\bin;%PATH%

[emkornfield]

@kou I made the change, lets hope that fixes CI (I think the other changes aren't bad so I'll leave them in if that is OK). I sent a follow-up e-mail to the mailing to see how we want to proceed with some of these ignored warnings.

[emkornfield]

Build failure is plasma (looks like it might be flaky? because its based on timestamps):

 # Test timestamps
t1 = time.time()
x, _, _ = create_object(plasma_client, 3, metadata_size=0, seal=False)
t2 = time.time()
l6 = plasma_client.list()

 assert math.floor(t1) <= l6[x]["create_time"] <= math.ceil(t2)

E assert 1551585976 <= 1551585975
E + where 1551585976 = (1551585976.0053494)
E + where = math.floor

[kou]

+1
arrow-array-test is stuck with MinGW 32bit but it's not related in this change.
https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779178/job/787pa80bsth5rb4d#L727

[xhochy]

@kou@emkornfield The 64bit builds now get stuck from time to time too on master: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779601/job/d1h0bfv5oyecs414

[emkornfield]

@kou do you have thoughts on what might be causing that? Could we maybe disable ming64 until we can figure it out?

[kou]

I don't know why... It's not reproduced on my local machine.
We can disable only arrow-array-test instead of disabling all MinGW 64bit build. I want to keep Apache Arrow 0.13.0 buildable with MinGW.

[emkornfield]

@kou is disabling something you can do? I'm currently working on another flaky test and probably won't be able to get to it tonight. (otherwise I can try to do it tomorrow)

[kou]

OK. I'll do later.

[kou]

Done: #3809

[emkornfield]

No description provided.

[kou]

Thanks!
Should we do the same thing for Zstd?

https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22762420/job/4xuv4mpnwvmq5byf#L391

C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc: In member function 'virtual int64_t arrow::util::ZSTDCodec::MaxCompressedLen(int64_t, const uint8_t*)':
C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc:232:38: error: conversion to 'size_t {aka unsigned int}' from 'int64_t {aka long long int}' may alter its value [-Werror=conversion]
return ZSTD_compressBound(input_len);
^
cc1plus.exe: all warnings being treated as errors

[emkornfield]

Yes on it.

[emkornfield]

Also, should/can these checks be turned on for linux builds?

[emkornfield]

or turned off for ming in the short term? I don't understand why the breakage just started occurring now.

[kou]

It's caused by combination of #3693 and #3750. It's my fault. Sorry.
We can disable -Werror by the following change:

diff --git a/ci/appveyor-cpp-build-mingw.bat b/ci/appveyor-cpp-build-mingw.bat
index 249ea1ab..bdd3b144 100644
--- a/ci/appveyor-cpp-build-mingw.bat+++ b/ci/appveyor-cpp-build-mingw.bat@@ -17,8 +17,8 @@
@echo on
-set CMAKE_BUILD_TYPE=debug-set MESON_BUILD_TYPE=debug+set CMAKE_BUILD_TYPE=release+set MESON_BUILD_TYPE=release
set INSTALL_DIR=%HOMEDRIVE%%HOMEPATH%\install
set PATH=%INSTALL_DIR%\bin;%PATH%

[emkornfield]

@kou I made the change, lets hope that fixes CI (I think the other changes aren't bad so I'll leave them in if that is OK). I sent a follow-up e-mail to the mailing to see how we want to proceed with some of these ignored warnings.

[emkornfield]

Build failure is plasma (looks like it might be flaky? because its based on timestamps):

 # Test timestamps
t1 = time.time()
x, _, _ = create_object(plasma_client, 3, metadata_size=0, seal=False)
t2 = time.time()
l6 = plasma_client.list()

 assert math.floor(t1) <= l6[x]["create_time"] <= math.ceil(t2)

E assert 1551585976 <= 1551585975
E + where 1551585976 = (1551585976.0053494)
E + where = math.floor

[kou]

+1
arrow-array-test is stuck with MinGW 32bit but it's not related in this change.
https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779178/job/787pa80bsth5rb4d#L727

[xhochy]

@kou@emkornfield The 64bit builds now get stuck from time to time too on master: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779601/job/d1h0bfv5oyecs414

[emkornfield]

@kou do you have thoughts on what might be causing that? Could we maybe disable ming64 until we can figure it out?

[kou]

I don't know why... It's not reproduced on my local machine.
We can disable only arrow-array-test instead of disabling all MinGW 64bit build. I want to keep Apache Arrow 0.13.0 buildable with MinGW.

[emkornfield]

@kou is disabling something you can do? I'm currently working on another flaky test and probably won't be able to get to it tonight. (otherwise I can try to do it tomorrow)

[kou]

OK. I'll do later.

[kou]

Done: #3809

[emkornfield]

No description provided.

[kou]

Thanks!
Should we do the same thing for Zstd?

https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22762420/job/4xuv4mpnwvmq5byf#L391

C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc: In member function 'virtual int64_t arrow::util::ZSTDCodec::MaxCompressedLen(int64_t, const uint8_t*)':
C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc:232:38: error: conversion to 'size_t {aka unsigned int}' from 'int64_t {aka long long int}' may alter its value [-Werror=conversion]
return ZSTD_compressBound(input_len);
^
cc1plus.exe: all warnings being treated as errors

[emkornfield]

Yes on it.

[emkornfield]

Also, should/can these checks be turned on for linux builds?

[emkornfield]

or turned off for ming in the short term? I don't understand why the breakage just started occurring now.

[kou]

It's caused by combination of #3693 and #3750. It's my fault. Sorry.
We can disable -Werror by the following change:

diff --git a/ci/appveyor-cpp-build-mingw.bat b/ci/appveyor-cpp-build-mingw.bat
index 249ea1ab..bdd3b144 100644
--- a/ci/appveyor-cpp-build-mingw.bat+++ b/ci/appveyor-cpp-build-mingw.bat@@ -17,8 +17,8 @@
@echo on
-set CMAKE_BUILD_TYPE=debug-set MESON_BUILD_TYPE=debug+set CMAKE_BUILD_TYPE=release+set MESON_BUILD_TYPE=release
set INSTALL_DIR=%HOMEDRIVE%%HOMEPATH%\install
set PATH=%INSTALL_DIR%\bin;%PATH%

[emkornfield]

@kou I made the change, lets hope that fixes CI (I think the other changes aren't bad so I'll leave them in if that is OK). I sent a follow-up e-mail to the mailing to see how we want to proceed with some of these ignored warnings.

[emkornfield]

Build failure is plasma (looks like it might be flaky? because its based on timestamps):

 # Test timestamps
t1 = time.time()
x, _, _ = create_object(plasma_client, 3, metadata_size=0, seal=False)
t2 = time.time()
l6 = plasma_client.list()

 assert math.floor(t1) <= l6[x]["create_time"] <= math.ceil(t2)

E assert 1551585976 <= 1551585975
E + where 1551585976 = (1551585976.0053494)
E + where = math.floor

[kou]

+1
arrow-array-test is stuck with MinGW 32bit but it's not related in this change.
https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779178/job/787pa80bsth5rb4d#L727

[xhochy]

@kou@emkornfield The 64bit builds now get stuck from time to time too on master: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779601/job/d1h0bfv5oyecs414

[emkornfield]

@kou do you have thoughts on what might be causing that? Could we maybe disable ming64 until we can figure it out?

[kou]

I don't know why... It's not reproduced on my local machine.
We can disable only arrow-array-test instead of disabling all MinGW 64bit build. I want to keep Apache Arrow 0.13.0 buildable with MinGW.

[emkornfield]

@kou is disabling something you can do? I'm currently working on another flaky test and probably won't be able to get to it tonight. (otherwise I can try to do it tomorrow)

[kou]

OK. I'll do later.

[kou]

Done: #3809

[emkornfield]

No description provided.

[kou]

Thanks!
Should we do the same thing for Zstd?

https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22762420/job/4xuv4mpnwvmq5byf#L391

C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc: In member function 'virtual int64_t arrow::util::ZSTDCodec::MaxCompressedLen(int64_t, const uint8_t*)':
C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc:232:38: error: conversion to 'size_t {aka unsigned int}' from 'int64_t {aka long long int}' may alter its value [-Werror=conversion]
return ZSTD_compressBound(input_len);
^
cc1plus.exe: all warnings being treated as errors

[emkornfield]

Yes on it.

[emkornfield]

Also, should/can these checks be turned on for linux builds?

[emkornfield]

or turned off for ming in the short term? I don't understand why the breakage just started occurring now.

[kou]

It's caused by combination of #3693 and #3750. It's my fault. Sorry.
We can disable -Werror by the following change:

diff --git a/ci/appveyor-cpp-build-mingw.bat b/ci/appveyor-cpp-build-mingw.bat
index 249ea1ab..bdd3b144 100644
--- a/ci/appveyor-cpp-build-mingw.bat+++ b/ci/appveyor-cpp-build-mingw.bat@@ -17,8 +17,8 @@
@echo on
-set CMAKE_BUILD_TYPE=debug-set MESON_BUILD_TYPE=debug+set CMAKE_BUILD_TYPE=release+set MESON_BUILD_TYPE=release
set INSTALL_DIR=%HOMEDRIVE%%HOMEPATH%\install
set PATH=%INSTALL_DIR%\bin;%PATH%

[emkornfield]

@kou I made the change, lets hope that fixes CI (I think the other changes aren't bad so I'll leave them in if that is OK). I sent a follow-up e-mail to the mailing to see how we want to proceed with some of these ignored warnings.

[emkornfield]

Build failure is plasma (looks like it might be flaky? because its based on timestamps):

 # Test timestamps
t1 = time.time()
x, _, _ = create_object(plasma_client, 3, metadata_size=0, seal=False)
t2 = time.time()
l6 = plasma_client.list()

 assert math.floor(t1) <= l6[x]["create_time"] <= math.ceil(t2)

E assert 1551585976 <= 1551585975
E + where 1551585976 = (1551585976.0053494)
E + where = math.floor

[kou]

+1
arrow-array-test is stuck with MinGW 32bit but it's not related in this change.
https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779178/job/787pa80bsth5rb4d#L727

[xhochy]

@kou@emkornfield The 64bit builds now get stuck from time to time too on master: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779601/job/d1h0bfv5oyecs414

[emkornfield]

@kou do you have thoughts on what might be causing that? Could we maybe disable ming64 until we can figure it out?

[kou]

I don't know why... It's not reproduced on my local machine.
We can disable only arrow-array-test instead of disabling all MinGW 64bit build. I want to keep Apache Arrow 0.13.0 buildable with MinGW.

[emkornfield]

@kou is disabling something you can do? I'm currently working on another flaky test and probably won't be able to get to it tonight. (otherwise I can try to do it tomorrow)

[kou]

OK. I'll do later.

[kou]

Done: #3809

[emkornfield]

No description provided.

[kou]

Thanks!
Should we do the same thing for Zstd?

https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22762420/job/4xuv4mpnwvmq5byf#L391

C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc: In member function 'virtual int64_t arrow::util::ZSTDCodec::MaxCompressedLen(int64_t, const uint8_t*)':
C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc:232:38: error: conversion to 'size_t {aka unsigned int}' from 'int64_t {aka long long int}' may alter its value [-Werror=conversion]
return ZSTD_compressBound(input_len);
^
cc1plus.exe: all warnings being treated as errors

[emkornfield]

Yes on it.

[emkornfield]

Also, should/can these checks be turned on for linux builds?

[emkornfield]

or turned off for ming in the short term? I don't understand why the breakage just started occurring now.

[kou]

It's caused by combination of #3693 and #3750. It's my fault. Sorry.
We can disable -Werror by the following change:

diff --git a/ci/appveyor-cpp-build-mingw.bat b/ci/appveyor-cpp-build-mingw.bat
index 249ea1ab..bdd3b144 100644
--- a/ci/appveyor-cpp-build-mingw.bat+++ b/ci/appveyor-cpp-build-mingw.bat@@ -17,8 +17,8 @@
@echo on
-set CMAKE_BUILD_TYPE=debug-set MESON_BUILD_TYPE=debug+set CMAKE_BUILD_TYPE=release+set MESON_BUILD_TYPE=release
set INSTALL_DIR=%HOMEDRIVE%%HOMEPATH%\install
set PATH=%INSTALL_DIR%\bin;%PATH%

[emkornfield]

@kou I made the change, lets hope that fixes CI (I think the other changes aren't bad so I'll leave them in if that is OK). I sent a follow-up e-mail to the mailing to see how we want to proceed with some of these ignored warnings.

[emkornfield]

Build failure is plasma (looks like it might be flaky? because its based on timestamps):

 # Test timestamps
t1 = time.time()
x, _, _ = create_object(plasma_client, 3, metadata_size=0, seal=False)
t2 = time.time()
l6 = plasma_client.list()

 assert math.floor(t1) <= l6[x]["create_time"] <= math.ceil(t2)

E assert 1551585976 <= 1551585975
E + where 1551585976 = (1551585976.0053494)
E + where = math.floor

[kou]

+1
arrow-array-test is stuck with MinGW 32bit but it's not related in this change.
https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779178/job/787pa80bsth5rb4d#L727

[xhochy]

@kou@emkornfield The 64bit builds now get stuck from time to time too on master: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779601/job/d1h0bfv5oyecs414

[emkornfield]

@kou do you have thoughts on what might be causing that? Could we maybe disable ming64 until we can figure it out?

[kou]

I don't know why... It's not reproduced on my local machine.
We can disable only arrow-array-test instead of disabling all MinGW 64bit build. I want to keep Apache Arrow 0.13.0 buildable with MinGW.

[emkornfield]

@kou is disabling something you can do? I'm currently working on another flaky test and probably won't be able to get to it tonight. (otherwise I can try to do it tomorrow)

[kou]

OK. I'll do later.

[kou]

Done: #3809

[emkornfield]

No description provided.

[kou]

Thanks!
Should we do the same thing for Zstd?

https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22762420/job/4xuv4mpnwvmq5byf#L391

C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc: In member function 'virtual int64_t arrow::util::ZSTDCodec::MaxCompressedLen(int64_t, const uint8_t*)':
C:/projects/arrow/cpp/src/arrow/util/compression_zstd.cc:232:38: error: conversion to 'size_t {aka unsigned int}' from 'int64_t {aka long long int}' may alter its value [-Werror=conversion]
return ZSTD_compressBound(input_len);
^
cc1plus.exe: all warnings being treated as errors

[emkornfield]

Yes on it.

[emkornfield]

Also, should/can these checks be turned on for linux builds?

[emkornfield]

or turned off for ming in the short term? I don't understand why the breakage just started occurring now.

[kou]

It's caused by combination of #3693 and #3750. It's my fault. Sorry.
We can disable -Werror by the following change:

diff --git a/ci/appveyor-cpp-build-mingw.bat b/ci/appveyor-cpp-build-mingw.bat
index 249ea1ab..bdd3b144 100644
--- a/ci/appveyor-cpp-build-mingw.bat+++ b/ci/appveyor-cpp-build-mingw.bat@@ -17,8 +17,8 @@
@echo on
-set CMAKE_BUILD_TYPE=debug-set MESON_BUILD_TYPE=debug+set CMAKE_BUILD_TYPE=release+set MESON_BUILD_TYPE=release
set INSTALL_DIR=%HOMEDRIVE%%HOMEPATH%\install
set PATH=%INSTALL_DIR%\bin;%PATH%

[emkornfield]

@kou I made the change, lets hope that fixes CI (I think the other changes aren't bad so I'll leave them in if that is OK). I sent a follow-up e-mail to the mailing to see how we want to proceed with some of these ignored warnings.

[emkornfield]

Build failure is plasma (looks like it might be flaky? because its based on timestamps):

 # Test timestamps
t1 = time.time()
x, _, _ = create_object(plasma_client, 3, metadata_size=0, seal=False)
t2 = time.time()
l6 = plasma_client.list()

 assert math.floor(t1) <= l6[x]["create_time"] <= math.ceil(t2)

E assert 1551585976 <= 1551585975
E + where 1551585976 = (1551585976.0053494)
E + where = math.floor

[kou]

+1
arrow-array-test is stuck with MinGW 32bit but it's not related in this change.
https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779178/job/787pa80bsth5rb4d#L727

[xhochy]

@kou@emkornfield The 64bit builds now get stuck from time to time too on master: https://ci.appveyor.com/project/ApacheSoftwareFoundation/arrow/builds/22779601/job/d1h0bfv5oyecs414

[emkornfield]

@kou do you have thoughts on what might be causing that? Could we maybe disable ming64 until we can figure it out?

[kou]

I don't know why... It's not reproduced on my local machine.
We can disable only arrow-array-test instead of disabling all MinGW 64bit build. I want to keep Apache Arrow 0.13.0 buildable with MinGW.

[emkornfield]

@kou is disabling something you can do? I'm currently working on another flaky test and probably won't be able to get to it tonight. (otherwise I can try to do it tomorrow)

[kou]

OK. I'll do later.

[kou]

Done: #3809