patchMakefile: update defaults for modern Cygwin

Now that Cygwin 1.7.x has enabled lots of new features, and Cygwin 1.5 is no longer actively supported by the Cygwin mailing lists, we might as well update the defaults to cater to those new features.

NO_TRUSTABLE_FILEMODE is only necessary on FAT drives; the Cygwin community recommends NTFS drives, but there is still too much use for FAT to switch the default. Likewise, UNRELIABLE_FSTAT is probably file-system specific, but worth keeping unchanged.

This commit does not change the default for NO_MMAP, although definitive proof of whether this option is necessary is lacking.

Signed-off-by: Eric Blake <eblake@redhat.com>
---
 Makefile |   22 ++++++++++++----------
 1 files changed, 12 insertions(+), 10 deletions(-)

Eric Blake <eblake@redhat.com> writes:

Thanks; as I lack Cygwin environment to test this myself, I'll apply this directly on 'master' and see if anybody screams ;-)

Do we need to update this part when Cygwin 1.8 is released? Replacing this with "ifeq(... ,1 5)" may not be an improvement either, unless we are sure that nobody is using 1.4 or older, but I wonder if somebody else have better ideas?

On 04/02/2010 12:47 AM, Junio C Hamano wrote:

Thanks.

Odd numbers are public cygwin releases (1.5, 1.7), even numbers are Red Hat releases (1.6, 1.8), so cygwin 1.8 already does exist for Red Hat licensed customers. It will probably be years before cygwin 1.9/1.10 exists, at which point 1.5/1.6 will be such a distant memory that we could just delete this block altogether at that time.

Nothing short of using $(shell) and doing some actual computation on the string in $(uname_R). But that's not too hard, if people think it's worth it; speak up if you want me to rework it along those lines, otherwise I'll assume that things are good enough as-is.

Eric Blake wrote:

Red Hat internal people might want to be able to use 1.8 when it comes out. Maybe something like this would do the trick? I don’t know whether Cygwin 1.6 has the fixes 1.7 does, so in my ignorance I lumped it with 1.5.

What you sent is probably good enough already; just trying to avoid future work.

-- %< --
Subject: Makefile: future-proof Cygwin version check

Tweak the condition that detects old Cygwin versions to not include versions such as 1.8, 1.11, and 2.1.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
---

Thanks, will queue. By the way, please spell scissors as 8< or >8 not with a per-cent at least for now.

Handle perforations found “in the wild” more robustly by recognizing “%<” as an alternative scissors mark.

This feature is only meant to support old habits. Discourage new use of the percent-based version by only documenting the 8< symbol so new users’ perforations can still be recognized by old versions of Git.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
---
Junio C Hamano wrote:

Thanks for pointing it out.  I will from now on.

Searching for “cut here” on groups.google.com revealed several examples like this one:

 ______________________%<__cut here ______________>%_____________

[ed: dashes converted to underscores to avoid confusing am]

So I guess I’m the only one using the small-fingered scissors. Maybe this patch could help support old habits.

 builtin/mailinfo.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

On 04/03/2010 01:47 AM, Jonathan Nieder wrote:

Cygwin 1.6 is on par with 1.5 feature-wise, so your cutoff of 1.6 as the last old version is correct.

It would be nice to use fewer processes, since forking is so expensive on cygwin:

ifeq ($(shell case '$(uname_R)' in 1.[1-6].*) echo old;; esac),old)

But does that work to have unbalanced ) in a makefile $(shell)? On the other hand, this is already in a chunk guarded by $(uname_S) being cygwin, and we know that cygwin shells understand:

ifeq ($(shell case '$(uname_R)' in (1.[1-6].*) echo old;; esac),old)

even though it is not portable to other shells, like Solaris /bin/sh.

Eric Blake wrote:

Thanks for explaining.

Looks good to me.

While I have your attention, do you know of a simple way to test Cygwin programs under Linux? setup.exe does not work well under wine, so I am asking mostly in the hope that there is a .tar.gz or .zip binary distribution somewhere I could play with.

Thanks, Jonathan

On 04/05/2010 08:30 AM, Jonathan Nieder wrote:

Junio mentioned he had already queued your first version of the patch; are you going to re-submit it with this tweak?

Sadly, the fact that wine is not yet able to run setup.exe is evidence that wine is still lacking some emulation abilities, and while I am aware that the situation is trying to be improved, I am not actively participating in that effort. For now, I am only aware of the ability to run cygwin on native windows boxes (including virtual machines). In fact, I'm currently using a 240-day evaluation license of Windows Sever 2008 inside a VM to experiment with cygwin on my Linux box, without having spent any extra money.

From: Eric Blake <eblake@redhat.com>

It is nice to use fewer processes, since forking is so expensive on cygwin. So use a case statement instead of expr.

But does that work to have unbalanced ) in a makefile $(shell)? On the other hand, this is already in a chunk guarded by $(uname_S) being cygwin, and we know that cygwin shells understand:

	case '$(uname_R)' in (1.[1-6].*) echo old;; esac

even though it is not portable to other shells, like NetBSD ash.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
---
Eric Blake wrote:

[...]

Since the patch to future-proof the version check has already hit master, here’s the change as an independent patch against master.

 Makefile |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

Jonathan Nieder <jrnieder@gmail.com> writes:

Even less forks:

        ifneq ($(filter 1.1.% 1.2.% 1.3.% 1.4.% 1.5.% 1.6.%,$(uname_R)),)

Andreas.