Updated h5dump --xml by bmribler · Pull Request #6394 · HDFGroup/hdf5

bmribler · 2026-05-04T07:37:32Z

The location of the files HDF5-File.xsd and HDF5-File.dtd no longer exists, and the files were added to the repo (PR #5490). Before the --xml option is actually removed, running the h5dump --xml tests will fail. This PR updated h5dump and the expected output to use the correct files' location and actually completed the ticket HELP-2668.

The location of the files HDF5-File.xsd and HDF5-File.dtd no longer exists, and the files were added to the repo (PR HDFGroup#5490). Before the --xml is actually removed, running the h5dump tests will fail. This PR updated h5dump and the expected output to use the correct files' location and actually completed the ticket HELP-2668.

jhendersonHDF · 2026-05-04T19:39:25Z

Before the --xml option is actually removed, running the h5dump --xml tests will fail.

@bmribler what did you mean by this?

jhendersonHDF · 2026-05-04T19:40:02Z

@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="UTF-8"?>
-<hdf5:HDF5-File xmlns:hdf5="http://hdfgroup.org/HDF5/XML/schema/HDF5-File.xsd" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://hdfgroup.org/HDF5/XML/schema/HDF5-File http://www.hdfgroup.org/HDF5/XML/schema/HDF5-File.xsd">


At a first glance, it seems to me like these should remain as web URLs rather than local files

The files are now in the repo. Should I use the GH repo develop branch for the location?

As @brtnfld said in the standup, GH URL can be unreliable, but did we silently decide that we use it for now?

Yes — using the GitHub raw URL is the right approach since the files now live in the repo. That keeps the output as a proper web URL (addressing Jordan's concern) while pointing to a location that actually exists. Something like:

https://raw.githubusercontent.com/HDFGroup/hdf5/develop/tools/test/h5dump/testfiles/xml/HDF5-File.xsd

https://raw.githubusercontent.com/HDFGroup/hdf5/develop/tools/test/h5dump/testfiles/xml/HDF5-File.dtd

The ../../testfiles/xml/ relative path in the current PR bakes a test-directory layout assumption into the production binary output, which would break any downstream consumer that tries to resolve the schema location outside of the test tree.

jhendersonHDF · 2026-05-04T19:40:25Z

@@ -1597,8 +1597,8 @@ main(int argc, char *argv[])
                                "<%sHDF5-File xmlns:%s=\"http://hdfgroup.org/HDF5/XML/schema/HDF5-File.xsd\" "


But having said that, it seems this would need to be updated as well if changing the URLs

These are just names, the files won't be looked for by anything, if I'm not mistaken. I was thinking about using the GH repo URLs, but waited to ask when this is brought up.

This line sets the xmlns namespace URI (xmlns:hdf5="http://hdfgroup.org/HDF5/XML/schema/HDF5-File.xsd"), which is an opaque identifier per the XML Namespaces spec — it does not need to resolve to an actual resource. Changing it would alter the namespace identity of the output and silently break any downstream tool that identifies HDF5 XML output by its namespace URI, since namespace identity in XML is defined by that URI value, not by where the schema file lives. Only the xsi:schemaLocation location hint needs to point to the actual file. So this line should stay as-is.

Got it. Thanks!

The URL that this comment was made on still needs to be changed. You can probably just substitute in the DEFAULT_XSD string at this point.

It would likely be misuse of the namespace URI for downstream users to rely on it for anything, so I think it should be something less misleading since that URL doesn't point to anything useful and someone reading it may think it should.

bmribler · 2026-05-04T19:55:12Z

Before the --xml option is actually removed, running the h5dump --xml tests will fail.

@bmribler what did you mean by this?

@jhendersonHDF Thanks for looking at this. I'm not sure which part of that sentence you're asking about, so I explain the full context. --xml is deprecated, but before it is officially removed from h5dump, the tests should still work. Currently, because the location of those files no longer exist, tests will fail when those tests run. Please let me know if I have misunderstood anything.

jhendersonHDF · 2026-05-04T20:21:49Z

@jhendersonHDF Thanks for looking at this. I'm not sure which part of that sentence you're asking about, so I explain the full context. --xml is deprecated, but before it is officially removed from h5dump, the tests should still work. Currently, because the location of those files no longer exist, tests will fail when those tests run. Please let me know if I have misunderstood anything.

I was asking about the part mentioning those tests failing since we currently run the h5dump xml tests and they pass as they're only checking the diff output

bmribler · 2026-05-05T13:38:42Z

@jhendersonHDF Thanks for looking at this. I'm not sure which part of that sentence you're asking about, so I explain the full context. --xml is deprecated, but before it is officially removed from h5dump, the tests should still work. Currently, because the location of those files no longer exist, tests will fail when those tests run. Please let me know if I have misunderstood anything.

I was asking about the part mentioning those tests failing since we currently run the h5dump xml tests and they pass as they're only checking the diff output

Ah, ok. h5dump hardcoded the file names using the url to a non-existing location. I don't know when that matters in the tests, but it's definitely wrong.

brtnfld · 2026-06-03T22:14:01Z

Agreed it's wrong — the dead URL should be fixed. To clarify when it matters: the xsi:schemaLocation attribute is advisory for XSD-based files so most validators will silently ignore it if unreachable, but the DTD SYSTEM identifier (the <!DOCTYPE ... "url"> form) is normative — a validating XML parser will attempt to fetch it, so a broken URL there is a harder failure for anyone running validation downstream. Recommend updating DEFAULT_XSD and DEFAULT_DTD in h5dump.c to the GitHub raw URLs (see inline thread) and regenerating the fixtures, rather than the current local relative path.

bmribler · 2026-06-04T05:16:18Z

@brtnfld Yes, I fixed DEFAULT_XSD and DEFAULT_DTD in h5dump.c to the GH urls. It's just that I'm still missing something so having some test failures but I'll figure it out and push the changes soon.

… into update_xsdfile

bmribler · 2026-06-05T06:53:35Z

I made your change suggestions, @brtnfld

… into update_xsdfile

bmribler requested review from derobins, fortnern, jhendersonHDF, lrknox and qkoziol as code owners May 4, 2026 07:37

github-project-automation Bot added this to HDF5 - TRIAGE & TRACK May 4, 2026

bmribler requested review from brtnfld, glennsong09, mattjala and vchoi-hdfgroup as code owners May 4, 2026 07:37

github-project-automation Bot moved this to To be triaged in HDF5 - TRIAGE & TRACK May 4, 2026

Added missing expected files from the previous commit

ea38ac3

hyoklee previously approved these changes May 4, 2026

View reviewed changes

jhendersonHDF reviewed May 4, 2026

View reviewed changes

Fixed typo

625f16f

bmribler dismissed hyoklee’s stale review via 625f16f May 15, 2026 14:42

hyoklee previously approved these changes May 18, 2026

View reviewed changes

Merge branch 'develop' into update_xsdfile

7152621

lrknox assigned bmribler May 29, 2026

ajelenak added this to the HDF5 2.2.0 milestone Jun 2, 2026

bmribler added 2 commits June 5, 2026 02:43

Made corrections per review feedbacks

63d7454

Merge branch 'update_xsdfile' of https://github.com/bmribler/hdf5_bmr23…

3376cd6

… into update_xsdfile

bmribler dismissed hyoklee’s stale review via 3376cd6 June 5, 2026 06:51

Committing clang-format changes

5337e40

bmribler and others added 3 commits June 5, 2026 10:04

Merge branch 'develop' into update_xsdfile

ef4123b

Missing files from previous commit

c3e0ce7

Merge branch 'update_xsdfile' of https://github.com/bmribler/hdf5_bmr23…

b5fbd82

… into update_xsdfile

		@@ -1,5 +1,5 @@
		<?xml version="1.0" encoding="UTF-8"?>
		<hdf5:HDF5-File xmlns:hdf5="http://hdfgroup.org/HDF5/XML/schema/HDF5-File.xsd" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://hdfgroup.org/HDF5/XML/schema/HDF5-File http://www.hdfgroup.org/HDF5/XML/schema/HDF5-File.xsd">

		@@ -1597,8 +1597,8 @@ main(int argc, char *argv[])
		"<%sHDF5-File xmlns:%s=\"http://hdfgroup.org/HDF5/XML/schema/HDF5-File.xsd\" "

Uh oh!

Conversation

bmribler commented May 4, 2026

Uh oh!

jhendersonHDF commented May 4, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bmribler May 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

brtnfld Jun 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bmribler commented May 4, 2026

Uh oh!

jhendersonHDF commented May 4, 2026

Uh oh!

bmribler commented May 5, 2026

Uh oh!

brtnfld commented Jun 3, 2026

Uh oh!

bmribler commented Jun 4, 2026

Uh oh!

bmribler commented Jun 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

bmribler May 11, 2026 •

edited

Loading

brtnfld Jun 3, 2026 •

edited

Loading

bmribler commented Jun 5, 2026 •

edited

Loading