Personally, I am not a fan of unofficial urls.
If you want to download from repo, isn't it better to use vcpkg_from_git and point to https://bitbucket.hdfgroup.org/projects/HDFFV/repos/hdf5/browse ?

@cenit: give me a working example for vcpkg_from_git for hdf5 with the tag and I will change it. Otherwise I will not spend time on trying to get vcpkg_from_git to work ;)

vcpkg_from_git is working, no need to touch it

I put it as a suggestion in the code review panel, copied also here

vcpkg_from_git(
    OUT_SOURCE_PATH SOURCE_PATH
    URL https://bitbucket.hdfgroup.org/scm/hdffv/hdf5.git
    REF 83385326ef1588167f2b37300313a6e51e51ffca
)

(there it's missing the close bracket in order to continue with your PATCHES keyword

@cenit: You actually tried the code?

Starting package 3/3: hdf5:x64-windows-static
Building package hdf5[core,szip,zlib]:x64-windows-static...
-- Fetching https://bitbucket.hdfgroup.org/scm/hdffv/hdf5.git...
CMake Error at scripts/cmake/vcpkg_execute_required_process.cmake:72 (message):
    Command failed: "C:/Program Files/Git/mingw64/bin/git.exe" fetch https://bitbucket.hdfgroup.org/scm/hdffv/hdf5.git 83385326ef1588167f2b37300313a6e51e51ffca --depth 1 -n
    Working Directory: D:/vcpkg_common/downloads/git-tmp
    Error code: 128
    See logs for more information:
      D:\qt2\buildtrees\hdf5\git-fetch-x64-windows-static-err.log

fatal: expected shallow/unshallow, got The client mentioned a shallow object that is missing on the server. This may be due to a rebase or other forced update on the server. You should recreate your shallow clone.
fatal: the remote end hung up unexpectedly

Working on my Pc
Anyway, feel free to do what you prefer. Just stating my opinion about unofficial repositories...

@NancyLi1013: Can i assume that none of those regressions have anything to do with the changes made in this PR?

@Neumann-A
Thanks for your kindly reminder. I will rerun this PR and confirm if those regression can be solved.
Currently, I'm not sure how the regression for field3d on x64-linux is caused.

/mnt/vcpkg-ci/buildtrees/field3d/src/v1.7.2-ae08f6d321/src/Field3DFileHDF5.cpp:1138:71: error: too few arguments to function ‘herr_t H5Oget_info_by_name3(hid_t, const char*, H5O_info2_t*, unsigned int, hid_t)’
   status = H5Oget_info_by_name(loc_id, itemName, &infobuf, H5P_DEFAULT);
                                                                       ^
In file included from /mnt/vcpkg-ci/install/x64-linux/include/H5Apublic.h:22:0,
                 from /mnt/vcpkg-ci/install/x64-linux/include/hdf5.h:23,
                 from /mnt/vcpkg-ci/buildtrees/field3d/src/v1.7.2-ae08f6d321/src/Field3DFileHDF5.cpp:50:
/mnt/vcpkg-ci/install/x64-linux/include/H5Opublic.h:188:15: note: declared here
 H5_DLL herr_t H5Oget_info_by_name3(hid_t loc_id, const char *name, H5O_info2_t *oinfo,
               ^~~~~~~~~~~~~~~~~~~~
/mnt/vcpkg-ci/buildtrees/field3d/src/v1.7.2-ae08f6d321/src/Field3DFileHDF5.cpp: In function ‘herr_t Field3D::v1_7::InputFileHDF5::parseLayers(hid_t, const char*, const H5L_info2_t*, void*)’:

   status = H5Oget_info_by_name (loc_id, itemName, &infobuf, H5P_DEFAULT);

/azp run

@Neumann-A
Seems the regression is related with this hdf5.
Could you please look into this?

@NancyLi1013: TODO for the future: Fix field3d hidden MPI dependency

/azp run

@NancyLi1013: TODO for the future: Fix field3d hidden MPI dependency

It’s already almost done in my opencv 4.3 pr

@cenit So, should we waiting for your PR #11130?

yes please. I just pushed the patch for field3d there.
Waiting for CI results, hoping for that PR to be done...

@cenit
Thanks for your attention and kindly reminder. Actually, we always recommend to use official repos except for some specific reasons. And now I noticed that @Neumann-A has updated as official repo for hdf5 in this PR. Really appreciate.

@Neumann-A
Is there anything else to be done for this PR? If not, I will try to test these features.

@NancyLi1013: Nothing to add. Feel free to resolve the conflict with the baseline.

Can you merge the latest changes into this branch and use this to replace ci.baseline.txt?

@JackBoosY: I cleaned the baseline. If you need more changes than that those should probably be in another PR.

All features have passed with the following triplets:

• x86-windows
• x64-windows
• x64-windows-static

Note: Feature fortran has not been supported yet.

@JackBoosY @NancyLi1013´: Is there something blocking the merge of this PR?

@Neumann-A See #11747 (comment)

Waiting for merge #11130.

LGTM, just waiting for #11130 as suggested above. @Neumann-A Let us know if it makes more sense to merge this PR first.

@dan-shaw: seems like #11130 is still not finished so merging this one first seems more reasonable

#11130 is finished...

This change broke a lot of stuff :(

@BillyONeal. Instead of patching the sources use the following in the CMakeLists.txt:

add_definitions(-UH5_USE_112_API_DEFAULT)
add_definitions(-DH5_USE_110_API_DEFAULT)

there is probably also a way to set those via an environment variable from the portfile or using VCPKG_CXX_FLAGS