-
Notifications
You must be signed in to change notification settings - Fork 6.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[freetds] Update to v1.2.5 and update source to github. #14120
Conversation
Fixes failures in CI due to freetds.org being gone.
@freddy77 Do you have any comments on this? |
da1efb2
to
17ea1ee
Compare
It seems to me the final result is worst than before. They seem just some quick hacks to make it compile. Nothing wrong as temporary fix but I would look for a better final solution |
Can you be more specific on how it is worse?
Yes, that's what our patches are in the business of doing :)
Right, that's why I tagged you here since you seem to be the most recent contributor to the upstream repo and it would be nice to distil this down into something you might actually want to merge there for these 2 changes (adding an option to skip building the unit tests, and making the cmake build system actually correct with respect to encoding.h) |
Yes,
What's wrong with |
That's how CMake does this. If you don't make it a target then there is no dependency ordering so nothing forces that
That is intentional; libraries' users don't need or care about them: https://github.com/microsoft/vcpkg/blob/master/docs/maintainers/maintainer-guide.md#do-not-build-testsdocsexamples-by-default
Pool part?
Yes, if you have enough cores, cmake tries to build things that need encoding.h before encoding.h is created in the upstream build system. That doesn't happen with the tar.gz because there you have minted encoding.h in advance. |
Understood. Trying something more complete, see FreeTDS/freetds@f835b99
Mumble... with Autoconf a simple
|
Ah, I see, you have EXCLUDE_FROM_ALL'd them, so that patch might indeed be unnecessary.
Ah, I did that because I think we are just deleting the output of those targets, both before and after this change; for example vcpkg/ports/freetds/portfile.cmake Line 44 in 180cae5
That seems to have made the problem worse, since now iconv_charsets.h isn't being treated as a dependency of the tds port after that change since it isn't generated under that current directory. On my 64 thread system with vcpkg targeted at that commit SHA and all patches removed now I get:
[,..]
|
Not that would kill compiling a couple of more files.
Fixed, my mistake, see FreeTDS/freetds@d336746 |
Fixes failures in CI due to freetds.org being gone.