25 Aug
2021
25 Aug
'21
9:24 p.m.
Further simplify get_file_info by using fstatat for both path-based
(lstat) and fd-based (fstat) modes of operation.
Best,
Erich
29 Aug
29 Aug
4:53 a.m.
This sadly won't work for macOS 10.8 or 10.9. fstatat(2) was introduced to macOS in
10.10.
is_mount_point() already has this problem, but that's not your fault.
Chip
5:18 a.m.
Are you referring to a runtime issue?, since a later SDK has been needed
from wine-5.8.
On Sun, Aug 29, 2021 at 12:54 AM Chip Davis <cdavis(a)codeweavers.com> wrote:
This sadly won't work for macOS 10.8 or 10.9.
fstatat(2) was introduced to
macOS in 10.10.
is_mount_point() already has this problem, but that's not your fault.
Chip
5:40 a.m.
August 29, 2021 12:18 AM, "Dean Greer" <gcenx83(a)gmail.com
(mailto:[email protected]?to=%22Dean%20Greer%22%20<[email protected]>)> wrote:
Are you referring to a runtime issue?, since a later SDK has been needed from wine-5.8.
Yes. That function will be weak-linked, which means that on 10.8 and 10.9 it will be NULL.
Calling it without checking for NULL will cause a crash on those systems.
On Sun, Aug 29, 2021 at 12:54 AM Chip Davis <cdavis(a)codeweavers.com
(mailto:[email protected])> wrote:This sadly won't work for macOS 10.8 or
10.9. fstatat(2) was introduced to macOS in 10.10.
is_mount_point() already has this problem, but that's not your fault.
Chip
Chip
30 Aug
30 Aug
7:32 p.m.
On Sat, Aug 28, 2021 at 11:40 PM Chip Davis <cdavis(a)codeweavers.com> wrote:
August 29, 2021 12:18 AM, "Dean Greer"
<gcenx83(a)gmail.com> wrote:
I was obviously expecting this to not be a problem since fstatat is
already used in fd_is_mount_point(), is there an existing example of
how we like to work around a function that's weak-linked and possibly
NULL?
Best,
Erich
Are you referring to a runtime issue?, since a
later SDK has been needed from wine-5.8.
Yes. That function will be weak-linked,
which means that on 10.8 and 10.9 it will be NULL. Calling it without checking for NULL
will cause a crash on those systems.
7:43 p.m.
August 30, 2021 2:32 PM, "Erich E. Hoover" <erich.e.hoover(a)gmail.com>
wrote:
On Sat, Aug 28, 2021 at 11:40 PM Chip Davis
<cdavis(a)codeweavers.com> wrote:
See lines 92-96 of dlls/ntdll/unix/sync.c. Though in this case, I don't think that
method will be acceptable here, since mach_continuous_time() is unique to macOS and is
used in macOS-specific code, while fstatat(2) is a standard POSIX function used in code
compiled everywhere.
Chip
August 29, 2021 12:18 AM, "Dean Greer"
<gcenx83(a)gmail.com> wrote:
Are you referring to a runtime issue?, since a later SDK has been needed from wine-5.8.
Yes. That function will be weak-linked, which means that on 10.8 and 10.9 it will be
NULL. Calling
it without checking for NULL will cause a crash on those systems.
I was obviously expecting this to not be a problem since fstatat is
already used in fd_is_mount_point(), is there an existing example of
how we like to work around a function that's weak-linked and possibly
NULL?
7:53 p.m.
How badly is this function needed as there is a third-party solution that
implements this and more missing functions.
On Mon, Aug 30, 2021 at 3:32 PM Erich E. Hoover <erich.e.hoover(a)gmail.com>
wrote:
On Sat, Aug 28, 2021 at 11:40 PM Chip Davis
<cdavis(a)codeweavers.com>
wrote:
August 29, 2021 12:18 AM, "Dean Greer"
<gcenx83(a)gmail.com> wrote:
> Are you referring to a runtime issue?, since a later SDK has been
needed from
wine-5.8.
Yes. That function will be weak-linked, which
means that on 10.8 and
10.9 it will be NULL. Calling it without checking for NULL
will cause a
crash on those systems.
I was obviously expecting this to not be a problem since fstatat is
already used in fd_is_mount_point(), is there an existing example of
how we like to work around a function that's weak-linked and possibly
NULL?
Best,
Erich
8:20 p.m.
On Mon, Aug 30, 2021 at 1:53 PM Dean Greer <gcenx83(a)gmail.com> wrote:
How badly is this function needed as there is a
third-party solution that implements this and more missing functions.
We're currently using (even before this patch) the fstatat() function
to detect if crossing a symlink puts us on a different drive. I'd
definitely prefer for this to be "not our problem" by either using a
library to fill in these routines or upping the minimum supported Mac
OS X version. That said, it's not that hard on Mac OS X to implement
a replacement for fstatat() using fcntl(F_GETPATH).
Best,
Erich
8:33 p.m.
Wine was using the function for a long time but was avoided on macOS up
until wine-5.8 when the old buildbot was taken out of commission.
With how much of a PITA it is setting up a build environment for compiling
wine on MacOSX10.8/10.9 I’d leave this as not our problem, it’s simply a
matter of using the replacement library to fill in the gaps avoiding even
needing a newer SDK.
On Mon, Aug 30, 2021 at 4:20 PM Erich E. Hoover <erich.e.hoover(a)gmail.com>
wrote:
On Mon, Aug 30, 2021 at 1:53 PM Dean Greer
<gcenx83(a)gmail.com> wrote:
How badly is this function needed as there is a
third-party solution
that implements this and more missing functions.
We're currently using (even before this patch) the fstatat() function
to detect if crossing a symlink puts us on a different drive. I'd
definitely prefer for this to be "not our problem" by either using a
library to fill in these routines or upping the minimum supported Mac
OS X version. That said, it's not that hard on Mac OS X to implement
a replacement for fstatat() using fcntl(F_GETPATH).
Best,
Erich
7:48 p.m.
On 8/28/21 11:53 PM, Chip Davis wrote:
This sadly won't work for macOS 10.8 or 10.9.
fstatat(2) was introduced to macOS in 10.10.
is_mount_point() already has this problem, but that's not your fault.
Gijs and Dean proposed in [1] to bump the minimum required macOS version
to 10.13. Is there a good reason to support versions older than 10.10?
[1] https://www.winehq.org/pipermail/wine-devel/2021-July/191504.html
7:54 p.m.
We’d proposed that for the prebuilt packages as it’s PITA to build
libraries that are compatible with the lower versions of macosx, however
the library I mentioned to you prior already covers this and more missing
functions.
On Mon, Aug 30, 2021 at 3:48 PM Zebediah Figura (she/her) <
zfigura(a)codeweavers.com> wrote:
On 8/28/21 11:53 PM, Chip Davis wrote:
This sadly won't work for macOS 10.8 or 10.9.
fstatat(2) was introduced
to macOS in 10.10.
is_mount_point() already has this problem, but that's not your fault.
Gijs and Dean proposed in [1] to bump the minimum required macOS version
to 10.13. Is there a good reason to support versions older than 10.10?
[1] https://www.winehq.org/pipermail/wine-devel/2021-July/191504.html
8:36 p.m.
On 8/30/21 2:54 PM, Dean Greer wrote:
We’d proposed that for the prebuilt packages as it’s
PITA to build
libraries that are compatible with the lower versions of macosx, however
the library I mentioned to you prior already covers this and more missing
functions.
Maybe I'm missing something, but compatibility libraries are a moot
point if this function is available in 10.13 and we're requiring 10.13,
right?
8:40 p.m.
Correct the official Winehq packages will be targeting 10.13 so for that
it’s a moot point.
The thing is if wine still “officially” supports down to 10.8 how to handle
the missing functions, I’d go with using a third party implementation that
would then make needing a newer SDK a moot point.
On Mon, Aug 30, 2021 at 4:36 PM Zebediah Figura (she/her) <
zfigura(a)codeweavers.com> wrote:
On 8/30/21 2:54 PM, Dean Greer wrote:
We’d proposed that for the prebuilt packages as
it’s PITA to build
libraries that are compatible with the lower versions of macosx, however
the library I mentioned to you prior already covers this and more missing
functions.
Maybe I'm missing something, but compatibility libraries are a moot
point if this function is available in 10.13 and we're requiring 10.13,
right?
20 Sep
20 Sep
6:18 p.m.
Chip,
Do we need to do something to deal with this or do you feel like I can
resend these patches?
Best,
Erich
On Mon, Aug 30, 2021 at 2:40 PM Dean Greer <gcenx83(a)gmail.com> wrote:
Correct the official Winehq packages will be targeting 10.13 so for that it’s a moot
point.
The thing is if wine still “officially” supports down to 10.8 how to handle the missing
functions, I’d go with using a third party implementation that would then make needing a
newer SDK a moot point.
On Mon, Aug 30, 2021 at 4:36 PM Zebediah Figura (she/her) <zfigura(a)codeweavers.com>
wrote:
>
> On 8/30/21 2:54 PM, Dean Greer wrote:
> > We’d proposed that for the prebuilt packages as it’s PITA to build
> > libraries that are compatible with the lower versions of macosx, however
> > the library I mentioned to you prior already covers this and more missing
> > functions.
>
> Maybe I'm missing something, but compatibility libraries are a moot
> point if this function is available in 10.13 and we're requiring 10.13,
> right?
6:29 p.m.
September 20, 2021 1:19 PM, "Erich E. Hoover" <erich.e.hoover(a)gmail.com>
wrote:
Chip,
Do we need to do something to deal with this or do you feel like I can
resend these patches?
Well, it won't break things any more than they already are, and Dean says we can build
with a compatibility layer to make this work on earlier versions. I guess nothing else
needs to be done on your part, then.
Best,
Erich
On Mon, Aug 30, 2021 at 2:40 PM Dean Greer <gcenx83(a)gmail.com> wrote:
> Correct the official Winehq packages will be targeting 10.13 so for that it’s a moot
point.
>
> The thing is if wine still “officially” supports down to 10.8 how to handle the
missing functions,
> I’d go with using a third party implementation that would then make needing a newer
SDK a moot
> point.
>
> On Mon, Aug 30, 2021 at 4:36 PM Zebediah Figura (she/her)
<zfigura(a)codeweavers.com> wrote:
>> On 8/30/21 2:54 PM, Dean Greer wrote:
>>> We’d proposed that for the prebuilt packages as it’s PITA to build
>>> libraries that are compatible with the lower versions of macosx, however
>>> the library I mentioned to you prior already covers this and more missing
>>> functions.
>>
>> Maybe I'm missing something, but compatibility libraries are a moot
>> point if this function is available in 10.13 and we're requiring 10.13,
>> right?
Chip
958
days inactive
984
days old
14 comments
4 participants
participants (4)
-
Chip Davis -
Dean Greer -
Erich E. Hoover -
Zebediah Figura (she/her)