Error loading modules: wrong file length

[The-city-not-present]

There is some issue. I'll try to locate the problem but I'll post here for visibility first.

Sometimes with some of the files this tool works fine. But sometimes it creates the bundle, but then when importing the package from the bundle it breaks with unexpected and different errors. It was very hard to understand what is going on, but then I realized the piece of code that is loaded for the module is not ended at the right place but it includes the beginning of the next file. I've edited importer.template for debugging and edited get_code and added this next to compiled_code = compile...: I added with open(filename+'.sample.py','w',encoding='utf-8') as f: f.write(code) f.close() Given that, I can now see the module code that is loaded, and I see that the file does not end at the right place but there is a beginning of a next file added. This does not happen always, this happens with some of the files.

[The-city-not-present]

I see why it happens. Everything is looking so simple. First we write to a file, we call f.seek(), f.tell(), and capturing character positions. But in Windows line break characters ar CRLF, 2 chars. Then, when loading code, we are calling TextIOWrapper.read that stores text with normalized line breaks, and every line break is just \n, and the total length is different.

This is still strange. Starting position is working correctly, and datafile.seek(start) is working correctly. But when reading the specified length the length is counted in normalized line breaks with \n, not CRLF.

So, when checking len(code) that was loaded it matches end-start. But it is counted with \n as a line break, and in fact it includes bigger piece.

The solution that works for me is to change how ending position is captured. So in pinliner.py I changed package_end = f.tell() to package_end = package_start + len(code)

[The-city-not-present]

Maybe it makes sense to rename package_end to length and keep capturing the length only, because what matters is the number of chars we read, not the position.

[The-city-not-present]

I tried to post a fix - to create a branch as described at "pull request beginner guide" and push to that branch. I was rejected with HTTP 403.

I created a fork but that fork includes some more changes. I was only trying to address the issue in pull request, with minimal changes, nothing else. I was not trying to push what I have in the fork.

[andreyputilovmaterial]

Actually I submitted a PR. This was rather unintentional. It includes code updates that I have in my forked version, addressing both issues #10 and #11.