java.io.File.canWrite() is not sufficient to test whether ".cpcache" dir can be created

[jdjohnston]

This is an issue on Windows 7. Can't say if it's true for newer Windows or other OSes.

I was looking for a test case to try verify that I was actually using 1.11.1.1429 by setting DEPS_CLJ_TOOLS_VERSION as you suggested. One new item is the handling of .cpcache if the cwd isn't writable. deps.clj uses java.io.File.canWrite(), but that doesn't always work, which is easy to demonstrate in bb.

I set C:\Users\JDJ\tmp\xper such that I didn't have write or dir creation privileges in xper. I wasn't surprised that 1.11.1.1413 had problems, but 1.11.1.1429 was supposed to handle that case. However, I got the same error. After inspecting the Clojure code, I tried the following in bb:

C:\Users\JDJ\tmp\xper
λ bb
Babashka v1.3.186 REPL.
Use :repl/quit or :repl/exit to quit the REPL.
Clojure rocks, Bash reaches.

user=> (-> (io/file ".") (.canWrite))
true
user=> (fs/create-dir ".cpache")
java.nio.file.AccessDeniedException: .cpache [at :2:1]
user=>

Please ignore the typo (".cpache"). :)

[borkdude]

I'm open to a PR which fixes this. Bonus points for a test.

[jdjohnston]

Not sure when I'd have a chance to write up a PR. Test in Linux is, of course, easy: mkdir DIR; chmod 0555 DIR. Windows: shrug

As for code, I think you already have it: babashka.fs/writable? Without using babashka.fs directly here's another bb snippet [*]:

user=> (import 'java.nio.file.Files)
java.nio.file.Files
user=> (-> "." io/file (.toPath) Files/isWritable)
false

Not sure if there are any corner cases, but this works for R/O dirs in both Windows 7 & Linux, doing some quick tests.

[*] Thanks so much for bb. It makes experimenting in the Clojureverse not only easy, but downright fun!

[borkdude]

So there is a difference between canWrite and isWriteable, that is the crux?

[jdjohnston]

Seems to be, based on some _lite_ testing on Windows 7 & Debian Linux. Not sure they're supposed to be different, but that's a Java issue

…

On Thu, Dec 21, 2023, 04:47 Michiel Borkent ***@***.***> wrote: So there is a difference between canWrite and isWriteable, that is the crux? — Reply to this email directly, view it on GitHub <#120 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD4V3RK7NBAP3IJG5DW3FATYKQARVAVCNFSM6AAAAABA5QWUNCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRVHE2TONJVGI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[jlesquembre]

@jdjohnston Can you provide an example where canWrite and isWriteable return a different value? I can't reproduce it.

[jdjohnston]

As my original Issue description says, this might be specific to Winsows, maybe even just older Windows. I used Windows Explorer to remove write & dir creation privileges from the test dir. In that dir there is consistently a difference. Sorry I can't provide a command-line sequence, but I had to use the GUI to remove the privileges

…

On Thu, Dec 21, 2023, 13:28 José Luis Lafuente ***@***.***> wrote: @jdjohnston <https://github.com/jdjohnston> Can you provide an example where canWrite and isWriteable return a different value? I can't reproduce it. — Reply to this email directly, view it on GitHub <#120 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD4V3RIX6GLQJ47ORPTZ4RLYKR5WVAVCNFSM6AAAAABA5QWUNCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRWG43DCNBYG4> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[jdjohnston]

I think I may have confused things earlier when I mentioned Debian Linux. I used Linux to verify that isWritable would give the desired result, _not_ that it gave a different result from canWrite. As one would hope, for a simple test at least, they give the same result on Linux. So sorry for any confusion! :( On Thu, Dec 21, 2023, 13:43 Jonathan D Johnston ***@***.***> wrote:

…

As my original Issue description says, this might be specific to Winsows, maybe even just older Windows. I used Windows Explorer to remove write & dir creation privileges from the test dir. In that dir there is consistently a difference. Sorry I can't provide a command-line sequence, but I had to use the GUI to remove the privileges On Thu, Dec 21, 2023, 13:28 José Luis Lafuente ***@***.***> wrote: > @jdjohnston <https://github.com/jdjohnston> Can you provide an example > where canWrite and isWriteable return a different value? I can't > reproduce it. > > — > Reply to this email directly, view it on GitHub > <#120 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AD4V3RIX6GLQJ47ORPTZ4RLYKR5WVAVCNFSM6AAAAABA5QWUNCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRWG43DCNBYG4> > . > You are receiving this because you were mentioned.Message ID: > ***@***.***> >

[borkdude]

Is the summary that we should switch to java.nio isWritable and then we can close the issue? Happy to do this.

[jdjohnston]

Sounds good to me!

…

On Thu, Dec 21, 2023, 15:19 Michiel Borkent ***@***.***> wrote: Is the summary that we should switch to java.nio isWritable and then we can close the issue? Happy to do this. — Reply to this email directly, view it on GitHub <#120 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD4V3RLQXTJ327SPV3PZSHTYKSKWPAVCNFSM6AAAAABA5QWUNCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRWHA4DENBQGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[borkdude]

ok, pushed. you can perhaps test the Windows build artifact when CI finishes. I can make a new release tomorrow or so if that's necessary

[jdjohnston]

Just tried the build artifact on my R/O Win 7 dir. As hoped for, deps.clj used the cache_dir since the cwd wasn't writable. Is there another corner case? Couldn't say, but at least _this_ one is now accounted for. Thanks & Merry Christmas!

…

On Thu, Dec 21, 2023, 15:29 Michiel Borkent ***@***.***> wrote: ok, pushed. you can perhaps test the Windows build artifact when CI finishes. I can make a new release tomorrow or so if that's necessary — Reply to this email directly, view it on GitHub <#120 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD4V3RP7KH4YRS3B62EETBLYKSLZFAVCNFSM6AAAAABA5QWUNCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRWHA4TCMRSGQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[borkdude]

New release is available

[jdjohnston]

@borkdude The new release recognizes that the test dir is R/O, as desired. No obvious regression :)