Reduce code differences between HDF5 1.8 and 1.10

[byrnHDF]

Mostly comments and whitespace.
Other changes are initialization statements (includes hid_t H5I_INVALID return values on errors) and simple brackets.
The goal was to reduce diffs between 1.10 and 1.8.
The FD and TS files had more extensive changes to reorder the functions to more closely match the order of those in 1.10.

[lrknox]

If autogen.sh is added, we will also need a .gitignore file, which will ignore the files in autom4te.cache and others, notably Makefile.ins if we're going to full autogen.sh mode. Otherwise I'm not sure we should check in the Makefile.in changes for a newer version of automake which will revert to the older version when regenerated on jelly.

[byrnHDF]

I forgot to remove that!
OR
should we add it?

[derobins]

We decided that the 1.8 branch would not get the autotools changes that went to what became 1.10. autogen.sh and the updated configure options should not be propagated to 1.8 without more discussion.

[bmribler]

I'm just curious, why are these listed individual instead of using wildcard? Should I do something differently with the files?

[lrknox]

Allen used qutogen.sh on his fedora system because the newer autotools there failed trying to generate trace files with bin/reconfigure. I’ve run bin/reconfigure on jelly and the Makefil.ins will be unchanged. I’ll add it to his PR#38 shortly. That doesn’t answer your question, but these additions will go away. Larry From: bmribler <[email protected]> Sent: Thursday, October 15, 2020 2:24 PM To: HDFGroup/hdf5 <[email protected]> Cc: Larry Knox <[email protected]>; Review requested <[email protected]> Subject: Re: [HDFGroup/hdf5] Hdf5 1 8 (#38) @bmribler commented on this pull request.

________________________________ In c++/test/Makefile.in<#38 (comment)>:

- -rm -rf ./$(DEPDIR)

+ -rm -f ./$(DEPDIR)/dsets.Po + -rm -f ./$(DEPDIR)/h5cpputil.Po + -rm -f ./$(DEPDIR)/tarray.Po + -rm -f ./$(DEPDIR)/tattr.Po + -rm -f ./$(DEPDIR)/tcompound.Po + -rm -f ./$(DEPDIR)/tdspl.Po + -rm -f ./$(DEPDIR)/testhdf5.Po + -rm -f ./$(DEPDIR)/tfile.Po + -rm -f ./$(DEPDIR)/tfilter.Po + -rm -f ./$(DEPDIR)/th5s.Po + -rm -f ./$(DEPDIR)/tlinks.Po + -rm -f ./$(DEPDIR)/tobject.Po + -rm -f ./$(DEPDIR)/trefer.Po + -rm -f ./$(DEPDIR)/ttypes.Po + -rm -f ./$(DEPDIR)/tvlstr.Po I'm just curious, why are these listed individual instead of using wildcard? Should I do something differently with the files? — You are receiving this because your review was requested. Reply to this email directly, view it on GitHub<#38 (review)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADGMWCVYLRAMSSMRTVSJIATSK5D53ANCNFSM4SR3VH4A>.

[derobins]

Need to fix the autotools issues before this can be approved.

[byrnHDF]

autotools reverted and other fixes for VFD options

[derobins]

There is way too much stuff going on in this PR for me to do anything other than blindly click the 'approve' button, especially given how bad github is as a code review tool.

Problems:

• Too many 'ideas' in a single PR.
• Significant autotools churn. If we're going to keep 1.8 around we might want to consider bringing over the 1.10+ autotools changes.
• We need a code review tool that works well for large PRs. Can we wire up Fisheye/Crucible or something similar to handle this? It'd have to be the online version since our last install was wildly out of date.

[lrknox]

Autotools files have been restored to previous state; bin/reconfigure runs on jelly, changing only 2 trace lines in src/H5FDhdfs.c.
Test configurations with c++, fortran, fortran2003, ros3, libhdfs, szip, zlib and parallel (separate from c++) compile and pass all tests now.