Re: "Permission denied" from dired, but not from ido completion in tramp

[Philipp Middendorf]

Hi Michael and Adam,

I asked the admins here finally and got some new results, namely from calling 
nfs4_getfacl instead of just getfacl:

# file: /asap3/petra3/gpfs/p11/2024/data/11019260
A::OWNER@:rxtTnNcCoy
A::GROUP@:rxtncy
A::EVERYONE@:tncy
A:fdg:fs-dmgt:rwaDdxtTnNcCoy
A:fd:psgsrv:rxtTnNcCy
A:fdg:11019260-part:rxtnc
A:fdg:11019260-clbt:rxtnc
A:fdg:11019260-dmgt:rxtnc
A:fdg:p11staff:rxtncy
A:fdg:p11dmgt:rxtncy
A:fd:asapo:rxtnc

As you can see, it mentions 11019260-part, which is a group I'm in. Not sure 
how that relates to emacs' handling of ACLs and if that can be fixed.

I have asked this previously but didn't get a reply: Why does emacs have this 
additional permissions check at all? Why doesn't it simply always try to open 
the file and then output the error from the operating system?

Cheers
Philipp

> Michael Albinus <michael.albinus@gmx.de> hat am 16.06.2024 15:55 CEST 
> geschrieben:
> 
>  
> Philipp Middendorf <pmidden@mailbox.org> writes:
> 
> > Hi Adam, hi Michael,
> 
> Hi Philipp,
> 
> > ls -ld returns:
> >
> > dr-xr-x--- 6 fsdata fsdata 4096 Jun  7 13:00 
> > /asap3/petra3/gpfs/p11/2024/data/11019260
> 
> Interesting. It doesn't show any ACL or SELinux indication in its
> permissions string, like "dr-xr-x---+" or "dr-xr-x---." And the
> permissions are the same as 'stat' has returned, according to your debug
> buffer:
> 
> --8<---------------cut here---------------start------------->8---
> 08:06:14.997750 tramp-send-command (6) # tramp_stat_file_attributes 
> /asap3/petra3/gpfs/p11/2024/data/11019260/ 2>/dev/null; echo 
> tramp_exit_status $?
> 08:06:15.041234 tramp-wait-for-regexp (6) #
> (("‘/asap3/petra3/gpfs/p11/2024/data/11019260/’") 6 ("fsdata" . 26666) 
> ("fsdata" . 6666) 1718366406 1717758044 1717758044 4096 "dr-xr-x---" t 
> 11510012908 -1)
> tramp_exit_status 0
> ///e9cd6126f97f56eae6b23d28b13a86f7#$
> --8<---------------cut here---------------end--------------->8---
> 
> That's what Tramp confuses.
> 
> > And mount | grep asap3 returns:
> >
> > core1 on /asap3 type gpfs (rw,relatime)
> 
> Well, I guess gpfs is a non-standard file system type, isn't it?
> Wikipedia tells me, it is "General Parallel File System, brand name IBM
> Storage Scale and previously IBM Spectrum Scale". Aha.
> 
> 'stat' doesn't return proper information in this case. However, 'test' does:
> 
> --8<---------------cut here---------------start------------->8---
> 08:06:10.158264 tramp-send-command (6) # test -r 
> /asap3/petra3/gpfs/p11/2024/data/11019260/ 2>/dev/null; echo 
> tramp_exit_status $?
> 08:06:10.206211 tramp-wait-for-regexp (6) #
> tramp_exit_status 0
> ///e9cd6126f97f56eae6b23d28b13a86f7#$
> --8<---------------cut here---------------end--------------->8---
> 
> Exit code 0 means, that the directory is readable.
> 
> > Hope that helps.
> 
> Yep. I've pushed a fix to the Tramp repository, which should handle this
> case. Do you have a chance to install Tramp from git, and test it? You
> don't need the defalias anymore. Instead, set the new user option
> tramp-use-file-attributes to nil.
> 
> > Cheers
> > Philipp
> 
> Best regards, Michael.