I attempted to compile current SVN of SWMR branch inside Cygwin using both the CYGWIN and MINGW32 versions of GCC. MINGW32 fails, but perhaps no attempt has been made to support it. It probably needs some compiler flags to tell it to behave like Windows when deciding which files to #include.
CYGWIN failed the "ATOMIC" test on local disk NTFS/Windows 7.
To be useful, SWMR needs to run on file systems where this is an issue. So my question is: what is the expected path forward? Does this become a non-issue when journaling is integrated? Do we need file systems locks?
SWMR has not been tested on Windows. The existing SWMR test program is a shell script that has not been converted to a scheme usable on Windows. We've also not tested it on Cygwin or MinGW at this time.
I'd have to look more carefully at the code for the atomic test, but I would not be surprised if it failed on NTFS due to a POSIX-Win32 translation issue in the Cygwin libraries, even if it should pass.
Windows (and Cygwin) support is something we'll be investigating, but for right now we're focusing on true POSIX platforms as we move the feature into 1.10.0.
I attempted to compile current SVN of SWMR branch inside Cygwin using both
the CYGWIN and MINGW32 versions of GCC. MINGW32 fails, but perhaps no
attempt has been made to support it. It probably needs some compiler
flags to tell it to behave like Windows when deciding which files to #include.
CYGWIN failed the "ATOMIC" test on local disk NTFS/Windows 7.
To be useful, SWMR needs to run on file systems where this is an issue.
So my question is: what is the expected path forward? Does this become a
non-issue when journaling is integrated? Do we need file systems locks?
I'm sorry for creating some confusion by mixing empirical test results for a single platform with a general question about a feature. The question was not directed specifically at Windows. I've been testing the SWMR code on both Windows and Linux.
My question is this: given that the "atomic" test fails on most (Linux and Windows) configurations, what is the best long term solution? Implementation advice is appreciated.
/bbaker
···
-----Original Message-----
From: Hdf-forum [mailto:hdf-forum-bounces@lists.hdfgroup.org] On Behalf Of Dana Robinson
Sent: Friday, June 26, 2015 5:53 AM
To: HDF Users Discussion List
Subject: Re: [Hdf-forum] SWMR and Win32
Hi,
SWMR has not been tested on Windows. The existing SWMR test program is a shell script that has not been converted to a scheme usable on Windows. We've also not tested it on Cygwin or MinGW at this time.
I'd have to look more carefully at the code for the atomic test, but I would not be surprised if it failed on NTFS due to a POSIX-Win32 translation issue in the Cygwin libraries, even if it should pass.
Windows (and Cygwin) support is something we'll be investigating, but for right now we're focusing on true POSIX platforms as we move the feature into 1.10.0.
I attempted to compile current SVN of SWMR branch inside Cygwin using
both the CYGWIN and MINGW32 versions of GCC. MINGW32 fails, but
perhaps no attempt has been made to support it. It probably needs
some compiler flags to tell it to behave like Windows when deciding
which files to #include.
CYGWIN failed the "ATOMIC" test on local disk NTFS/Windows 7.
To be useful, SWMR needs to run on file systems where this is an issue.
So my question is: what is the expected path forward? Does this
become a non-issue when journaling is integrated? Do we need file systems locks?
