Help testing gtools package on OSX

[Mauricio Caceres]

I'm hoping someone is willing to help me test the latest version of my gtools package on a Mac; it has already been tested on Windows and Linux so I don't expect any issues, but I would like to be sure before submitting an update to SSC. To do this, please uninstall gtools and install the latest version:

cap noi net uninstall gtools
local github https://raw.githubusercontent.com
net install gtools, from(`github'/mcaceresb/stata-gtools/master/build/) replace
gtools

Please make sure it says version 1.10.1 is installed. Then, from a fresh Stata session, run

do https://raw.githubusercontent.com/mcaceresb/stata-gtools/master/build/gtools_tests.do

On my local machine this took about an hour and a half. It creates a log file in the current directory with the results; please attach the log here or send it to me. Thanks in advance to anyone willing to help!

[William Lisowski]

Added in edit: I set up a guest account on my MacBook and the testing proceeded past the point of failure in the log below. I will report back when it finishes.

Sorry to report that it did not work. I am running macOS Monterrey 12.6.1 ("OSX" is deprecated terminology) on an M2 MacBook Air (Apple Silicon).

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
      name:  gtools_tests
       log:  /Users/lisowskiw/Downloads/gtools_tests_macosx.log
  log type:  text
 opened on:   8 Dec 2022, 10:32:14


Start:        10:32:14  8 Dec 2022
Options:      dependencies basic_checks comparisons switches bench_test
OS:           MacOSX
Machine Type: Mac (Apple Silicon)

-------------------------------------
Basic unit-tests 10:32:33  8 Dec 2022
-------------------------------------
    test(passed): checks_gcontract, oncollision(error)
    test(passed): checks_isid, oncollision(error)
    test(passed): checks_duplicates, oncollision(error)
    test(passed): checks_toplevelsof, oncollision(error)
    test(failed): checks_levelsof, oncollision(error)

WARNING: 10:33:03  8 Dec 2022
tests ran with non-0 exit status

tests ran but Stata gave error code r(9)

Start: 10:32:14 8 Dec 2022
End: 10:33:03  8 Dec 2022
      name:  gtools_tests
       log:  /Users/lisowskiw/Downloads/gtools_tests_macosx.log
  log type:  text
 closed on:   8 Dec 2022, 10:33:03
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

[William Lisowski]

The test on my guest account went much further but apparently failed as well. The log is too large to be uploaded; I have mailed it to your gmail address given in the output of ssc describe gtools.

[Mauricio Caceres]

William Lisowski Thanks so much for the help! I'm pretty sure the first error is mostly harmless; this check also failed in Windows at first and the issue was a slight levelsof vs glevelsof rounding when converting doubles to strings. I think lowering the tolerance of the comparison slightly should fix that.

The second is a bit weird. I've uploaded a test file that checks whether that function fails just with that algorithm or with every one. (It still runs the rest of the tests.) Would you be willing to try again? This time from the develop branch:

do https://raw.githubusercontent.com/mcaceresb/stata-gtools/develop/build/gtools_tests.do

Thanks!

[William Lisowski]

New output in email to be sent in a moment. Compressed into a ZIP archive, let me know if that doesn't work for you for some reason.

[Mauricio Caceres]

William Lisowski Thank you so much for all your help! This is very useful. I was testing this using Stata 16, it looks like. With Stata 17 I get the same failure, but I believe this is a bug with Stata (I reported this a few months ago here in another thread, if curious).

However, the function that failed originally keeps failing, and I was not able to replicate it. Would you be willing to send me another log file? This time it would solely be for that particular function; I've set up the test file for that so you can re-run

do https://raw.githubusercontent.com/mcaceresb/stata-gtools/develop/build/gtools_tests.do

Of course if not then you've already helped me more than enough (and given I flag this function as "beta" and it passes the unit tests on my machines I'd be OK with the current state of things).

Thanks again!

PS: This should only take a few minutes to run, unlike the full set of unit tests.

[William Lisowski]

New output to be sent in a moment. Since reghdfe had to be installed, I got it from SSC.

I'm glad to help, but I'll soon be quitting for the night (GMT - 5 here), so the next test will be tomorrow morning. It's no problem at all, as a pastime it has a higher satisfaction than some of the posts I commonly write.

[Mauricio Caceres]

William Lisowski No rush! I am puzzled that the tests didn't fail this time around... It might be some strange corner case I suppose. Can you try again? Same command (updated so it looks at the rest of the tests for it as well; again should only be a few minutes). Not sure what I'll conclude if this run also has no errors tho.

[William Lisowski]

Just sent those results, but it too ran successfully.

[Mauricio Caceres]

William Lisowski Sorry, for all the debugging, but I think I figured out the issue: I forgot to set reghdfe as a dependency. I think you may not have had it installed the very first time you ran the tests. I've made a release with the current code and I'm gonna submit the update to SSC. Thanks for all your help!

[William Lisowski]

I think you may not have had it installed the very first time you ran the tests

That is correct, it is not installed on my personal account (post #2). And when I created a guest account, it had nothing installed in PLUS because on Stata for Mac each user has their own PLUS directory, and the guest account as a new user would have been empty except for the dependencies installed by gtools (post #3).