Adodbapi missing files #1445
Adodbapi missing files #1445
Conversation
Closes mhammondgh-1151.
…odbapi_missing_files
catch up
…32 into adodbapi_missing_files
When executing python setup.py --user, self.prefix will be set to None. Of course, you will get an error that None can't be os.join'd. Therefore, using distutils' own variable, which has the script install directory already resolved.
… number objects in Python 3
…seconds. Should fix mhammond#1385, fix mhammond#387
This directory is now added to the start of PATH for version 3.7 and earlier, and passed to os.add_dll_directory() on 3.8 and later. This will hopefully work around problems loading pywintypes.dll in various situations. Fixes mhammond#1432, fixes mhammond#1431
|
I think you did pick the sweet-spot for testing, particularly if the servers being hit are ones not really expecting more traffic. That said though, the reality is that I'm probably not going to work out how to run the more exhaustive tests regularly, so there's a bit of a risk they will not get run. However, we can iterate on that over time. The changes here look fine. We should get something in CHANGES.txt too, explaining a bit of the motivation etc. If you could add that and then let me know when you think it's ready to merge, I'll pull the pin! |
|
Okay. I added a short paragraph in CHANGES.txt and a bit to adodbapi/readme.txt. |
|
Awesome, thanks! |
This PR does much more than the title suggests.
With this PR, the pywin32/adodbapi subfolder becomes the definitive source for adodbapi. Therefore, the source code can never get out of sync again.
The changes here include:
Query: with the addition of my three online database servers, it is possible to run a full "--all" test from anywhere, but that adds about five minutes, per pass, to the unit testing for USA based users. I have, therefore set the default test to run on only a local Jet database. It is not a complete test, but takes only around three seconds. I am working on a subsystem which will spin up Vagrant virtual machines for building and testing pywin32. If that were in place, the long Internet delay in testing would go away, and --all tests would be viable. Did I pick the correct optimization for the present moment?
Explanation: Testing of adodbapi has always been problematical because of the need for test databases. Microsoft has been kind enough to grant me an open source developer's free account which includes the ability to run a small VM on the azure cloud. Using that, I have spun up an Ubuntu-based SQL Server instance that we can use for testing. I have also opened up access on non-standard ports for MySQL and PostgreSQL servers on small VMs that I fund myself in France and California. None of these machines have enough capacity to make it worthwhile for anyone to use them for anything useful, nor are they mission critical for other uses.