adacore / e3-testsuite Goto Github PK

Specifically, of typing.Protocol in a couple of places:

There doesn't seem to be any specifically defined minimum Python version in this repo, but libadalang/langkit/e3-core do specify 3.7, so I guess this would inherit from that?

I'm facing an issue on GitHub Actions, when running e3-testsuite on windows. The issue concerns the rlimit tool from e3-core, I get the following message:

rlimit: cannot spawn process (error 0x5)

Looks like it's a known issue:

Is there something I should set in the GHA runner to make it work?

Thanks!

Dear e3-testsuite maintainer (😉),

Our main testsuite's driver is inheriting from the AdaCoreLegacyTestDriver and overrides its self.default_encoding with UTF-8. This makes sense for our usecase, we do not deal with binary data but we do deal with UTF-8 characters, hence having properly rendered UTF-8 characters is better than having \x[0-F][0-F] strings in our outputs.

Unfortunately, this does not work well on windows, some of our tests fail with the following backtrace:

Traceback (most recent call last):
  File "C:\it\sbx\wave\x86_64-windows64\e3-for-testing\install\lib\site-packages\e3\testsuite\fragment.py", line 187, in run
    self.return_value = self.data(self.previous_values, self.slot)
  File "C:\it\sbx\wave\x86_64-windows64\e3-for-testing\install\lib\site-packages\e3\testsuite\driver\classic.py", line 478, in run_wrapper
    self.set_up()
  File "C:\it\sbx\wave\x86_64-windows64\e3-for-testing\install\lib\site-packages\e3\testsuite\driver\adacore.py", line 91, in set_up
    self.argv = self.get_script_command_line()
  File "C:\it\sbx\wave\x86_64-windows64\e3-for-testing\install\lib\site-packages\e3\testsuite\driver\adacore.py", line 170, in get_script_command_line
    for line in f:
  File "C:\it\sbx\wave\x86_64-windows64\e3-for-testing\install\lib\encodings\cp1252.py", line 23, in decode
    return codecs.charmap_decode(input,self.errors,decoding_table)[0]
UnicodeDecodeError: 'charmap' codec can't decode byte 0x81 in position 193: character maps to <undefined>

This is because the test.sh contains UTF-8 characters, but the AdaCoreLegacyTestDriver opens test.sh files using the platform's default encoding.

The following patch seems to make things work again:

diff --git "a/C:\\it\\local\\e3\\lib\\site-packages\\e3\\testsuite\\driver\\adacore.py" b/tmp/driver.py
index 8625a50c9..af3e01642 100644
--- "a/C:\\it\\local\\e3\\lib\\site-packages\\e3\\testsuite\\driver\\adacore.py"
+++ b/tmp/driver.py
@@ -164,7 +164,7 @@ class AdaCoreLegacyTestDriver(DiffTestDriver):
                 new_script.append(". $TEST_SUPPORT_DIR/support.sh")

             # Read all lines in the original test script
-            with open(self.script_file) as f:
+            with open(self.script_file, encoding=self.default_encoding) as f:
                 # Get rid of potential whitespaces and CR at the end of
                 # each line.
                 for line in f:
@@ -177,10 +177,10 @@ class AdaCoreLegacyTestDriver(DiffTestDriver):

             # Write the shell script and schedule its execution with "bash"
             new_script_filename = self.working_dir("__test.sh")
-            with open(new_script_filename, "w") as f:
+            with open(new_script_filename, "wb") as f:
                 for line in new_script:
-                    f.write(line)
-                    f.write("\n")
+                    f.write(line.encode(self.default_encoding))
+                    f.write(b"\n")
             return ["bash", new_script_filename]

         else:  # os-specific

Since I'm not sure I have a complete understanding of the situation, I thought I'd open an issue instead of a PR, let me know if you need more information :)

In the following code:

The pattern substitute for the replacement of the path is buggy.
If the path is at the start of the line, it will replace some part of the previous line.

The issue comes from the first part of the regex:
rb"[^ '\"]*"

I suggest replacing this with:

rb"\S*"

Which takes care of whitespaces including the start of the line.

The text put in the <system-out> node is already escaped for XML inclusion, but it needs to be stripped from control characters as well, such as �^[2J�^[H which can be used to print coloured output in terminal. (Observed in the gnatfuzz testsuite)