PR introduces Flake8 errors 😲

See test results for Failed build of commit 19010f29db

PR introduces Flake8 errors 😲

See test results for Failed build of commit 915c5f8752

Was the only issue with this pr the comInterface_sonscript stuff, which has already been merged to master, or was there another issue... something about certificate signing?
My personal view is that we should research switching to vs 2019 for 2019.3 as much as possible, with the aim of then upgrading Liblouis to an official release again. However I do know @feerrenrut has some concirns around making too many changes in this release. At very least I think we need to see it working in action before we can say how bigger change it really is, and then debate if we should merge it. I think the argument that vs 2017 is now no longer the current version and is harder to get is a valid one.

The following entries from the appveyor build log of a try snapshot are relevant:

[00:00:14] if(!$env:APPVEYOR_PULL_REQUEST_NUMBER) {
[00:00:14]  openssl enc -aes-256-cbc -d -pass pass:$env:encFileKey -in authenticode.pfx.enc -out authenticode.pfx
[00:00:14]  openssl enc -aes-256-cbc -d -pass pass:$env:encFileKey -in ssh_id_rsa.enc -out ssh_id_rsa
[00:00:14]  # Install ssh stuff.
[00:00:14]  copy ssh_id_rsa $env:userprofile\.ssh\id_rsa
[00:00:14] }
[00:00:14] 
[00:00:15] openssl : *** WARNING : deprecated key derivation used.
[00:00:15] At line:2 char:2
[00:00:15] +  openssl enc -aes-256-cbc -d -pass pass:$env:encFileKey -in authentic ...
[00:00:15] +  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[00:00:15]     + CategoryInfo          : NotSpecified: (*** WARNING : d...erivation used.:String) [], RemoteException
[00:00:15]     + FullyQualifiedErrorId : NativeCommandError
[00:00:15]  
[00:00:15] Usi
[00:00:15] ng -iter or -pbkdf2 would be better.
[00:00:15] 
[00:00:15] b
[00:00:15] ad decrypt
[00:00:15] 
[00:00:15] 2576
[00:00:15] 99
[00:00:15] 02
[00:00:15] 14
[00:00:15] 4:
[00:00:15] erro
[00:00:15] r:
[00:00:15] 06065
[00:00:15] 06
[00:00:15] 4:dig
[00:00:15] it
[00:00:15] al envel
[00:00:15] op
[00:00:15] e rou
[00:00:15] tin
[00:00:15] es
[00:00:15] :E
[00:00:15] VP
[00:00:15] _Decr
[00:00:15] yp
[00:00:15] tFi
[00:00:15] n
[00:00:15] al
[00:00:15] _e
[00:00:15] x:
[00:00:15] bad d
[00:00:15] ecrypt:crypto/evp/evp_
[00:00:15] enc.c:537:
[00:00:15] 
[00:00:15] 
[00:00:15] openssl : *** WARNING : deprecated key derivation used.
[00:00:15] At line:3 char:2
[00:00:15] +  openssl enc -aes-256-cbc -d -pass pass:$env:encFileKey -in ssh_id_rs ...
[00:00:15] +  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[00:00:15]     + CategoryInfo          : NotSpecified: (*** WARNING : d...erivation used.:String) [], RemoteException
[00:00:15]     + FullyQualifiedErrorId : NativeCommandError
[00:00:15]  
[00:00:15] Using -iter or -pbkdf2 would be better.
[00:00:15] bad decrypt
[00:00:15] 25769902144:error:06065064:digital envelope routines:EVP_DecryptFinal_ex:bad decrypt:crypto/evp/evp_enc.c:537:
[00:02:19] signtool sign /fd SHA256 /f appveyor\authenticode.pfx /t http://timestamp.digicert.com C:\projects\nvda\build\x86\IAccessible2proxy.dll
[00:02:20] SignTool Error: An error occurred while attempting to load the signing

[00:02:20] 	certificate from: appveyor\authenticode.pfx

[00:02:20] scons: *** Error 1

The flake8 issue can be fixed by changing a line in tests/lint/flake8.ini.
On line 43, change:

include,

to

include/*,

In this pr, the scons submodule has changed, and therefore git diff lists include/scons as being changed. However, it seems that flake8 was not correctly matching include to include/scons. Looks like they need to match fully. thus include/* will cover any changes to submodules in include now.
As this is a draft pr, I was unable to push a commit to address this.

The vs2017 appveyor image includes openssl 1.0.2J where as vs2019 appveyor image contains openssl 1.1.1a.
It seems there is a change in openssl that no longer allows decrypting the way we used to.
Am doing further research.

openssl 1.1.0 changed its digest algorithm for deriving the key from md5 to sha256. For now, we can supply the -md md5 argument in our openssl commands in appveyor.yaml.

openssl enc -md md5 -aes-256-cbc -d -pass pass:$env:encFileKey -in authenticode.pfx.enc -out authenticode.pfx
    openssl enc -md md5 -aes-256-cbc -d -pass pass:$env:encFileKey -in ssh_id_rsa.enc -out ssh_id_rsa

Reference: https://stackoverflow.com/questions/39637388/encryption-decryption-doesnt-work-well-between-two-different-openssl-versions

Thanks for finding all this out.

• Regarding flake8, it turns out that excludes tells flake8 not to traverse into a directory, but if a file is provided in a diff, it will still check that file unless it is explicitly excluded, either with the file name or with a wild card. For now, I only changed include into include/*. If we ever change something in miscDeps, I don't think that will cause a problem, as only miscDeps will be in the diff and that's already excluded. Other excludes won't be in the diff as they are already excluded at the git level.

• Openssl: I changed this and pushes another try build

Actually, before getting in a rush, it might be good to point out that @egli filed an issue with the idea to port liblouis to rust. In that case, we will have to use binary builds anyway. In liblouis/liblouis#704 (comment), I asked whether there is a time frame for this.

having said that, I don't think it will happen for the next few releases.

@michaelDCurran: re liblouis, @egli pointed out that liblouis in rust won't be there any time soon. It is more of a wish, currently.

hi
Appveyor updated Windows images update on November 8, 2019
The visual studio2019 version has been updated to 16.3.8

@feerrenrut and @michaelDCurran, as you started merging in pull requests for 2020.1, I think this is a candidate to deserve priority, as #10275 relies on it which would benefit from an extensive testing period.

@LeonarddeR Yes I agree, I'll take a look at this soon.

As there are no extraordinary problems resulting from NVDA 2019.3, I think it is now safe to merge this to master, in order to give it a good run before a beta.