Windows PATH variables parsed with no '\'

[okuma10]

On windows using a PATH variable for includes ends up with a string with no ''
Example:
In Makefile: -I$(VK_SDK_PATH)/include
When make is run: -IC:\VulkanSDK\1.3.216.0/include
In compile_commands.json: -IC:VulkanSDK1.3.216.0/include

[arminveres]

I just noticed this issue as well. Is the PR #126 going to be merged i.e. the problem fixed in the foreseeable future?
Thanks for working on it @jinyi-x

[okuma10]

@arminveres did the fix worked for you? Nothing on my end

[arminveres]

@arminveres did the fix worked for you? Nothing on my end

Hi, I didn't try. So I am guessing it didn't work for you?
I just tried his branch, but it didn't do anything for me either.

[okuma10]

yes windows Path variables are without their "\" . I suppose it's a simple correction since everything is extracted correctly it just skips adding that character. I'll try to look into it next week.

[jinyi-x]

This patch fixed this problem on my system (Windows 10 LTSC 2021 & Python 3.10).And I noticed that if I switch my python into python3.6, it seems doesn't need any patch.Maybe it has relation with python version?I will keep testing. 来自 Outlook<http://aka.ms/weboutlook>

…

________________________________ 发件人: Armin Veres ***@***.***> 发送时间: 2022年11月10日 20:38 收件人: nickdiego/compiledb ***@***.***> 抄送: jinyi-x ***@***.***>; Mention ***@***.***> 主题: Re: [nickdiego/compiledb] Windows PATH variables parsed with no '\' (Issue #124) I just noticed this issue as well. Is the PR #126<#126> going to be merged i.e. the problem fixed in the foreseeable future? Thanks for working on it @jinyi-x<https://github.com/jinyi-x> ― Reply to this email directly, view it on GitHub<#124 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AOP5V6ZRNTWCJOP6QIQE4KLWHYAZHANCNFSM57IONGYQ>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[okuma10]

So I looked into it today and it appears the issue is due to this module --> bashlex, used to parse the command output from make

as you can see the preprocesed variable has the correct string, but the output of the bashlex.parser.parse() function is without the \\ character

Edit:
Adding a quick check for each character in preprocessed and if finding a '\' character and adding another fixed the issue. Oddly the clangd lsp server in nvim is having issues with the vulkan dir. but all else appears to be working fine and the compile_commands.json looks correct.

[okuma10]

I think this issue is solved by the above comment, maybe author can apply better change, but I also found this, trying to solve why the vulkan header wasn't being detected by nvim's clangd lsp.

Do you guys still get the same compile_commands.json ...with the arguments having " ' argument ' " instead of " argument "? @jinyi-x @arminveres

[arminveres]

I think this issue is solved by the above comment, maybe author can apply better change, but I also found this, trying to solve why the vulkan header wasn't being detected by nvim's clangd lsp.

Do you guys still get the same compile_commands.json ...with the arguments having " ' argument ' " instead of " argument "? @jinyi-x @arminveres

Nicely done. It really solves it. I am also gettting the " ' ' " quotation. I haven't investigated it, I only ran a quick sed from inside neovim and now it works!

I think you should propose a PR for this.

[okuma10]

@arminveres I looked up an old compile_commands.json generated from compiledb and that didn't had the " ' ' ", so it might be something from the change. I'll take a look at it later and will try to find why it adds those quotation marks.

[okuma10]

@arminveres I looked it up and it appears the " ' ' " pattern is introduced by @jinyi-x 's solution of using repr(s).encode().decode() in the unescape(s) function at the end.
Not sure what exactly that function does as there are no comments, but on my end Windows10 Python 3.11 returning to the original output of s.encode().decode() was giving me errors that cannot read a byte at the end of the string. So I just returned s, and that appeared to work. Might be a good idea for @nickdiego to give his input on this.
The unescape() function is used in the parse_build_log() function at around line 140(on my end since I might have commented out prints above it)

[arminveres]

@okuma10 It seems to work, if you drop jinyi's solution and just add yours. I have not tried it on Linux, though.

[GitMensch]

#128 seems to also take this into account...