diff options
| author | Adam J. Stewart <ajstewart426@gmail.com> | 2018-06-12 12:33:47 -0500 |
|---|---|---|
| committer | scheibelp <scheibel1@llnl.gov> | 2018-06-12 10:33:47 -0700 |
| commit | 0f1a1ae94e9b5f6e147bf13dae44a3af8d46c8fa (patch) | |
| tree | 3401b07d427a4e17d1cd78bdd0980e9c1dd93ceb | |
| parent | 583af6ef4fbeabebcaac1899bd902075bc46fb12 (diff) | |
| download | spack-0f1a1ae94e9b5f6e147bf13dae44a3af8d46c8fa.tar.gz spack-0f1a1ae94e9b5f6e147bf13dae44a3af8d46c8fa.tar.bz2 spack-0f1a1ae94e9b5f6e147bf13dae44a3af8d46c8fa.tar.xz spack-0f1a1ae94e9b5f6e147bf13dae44a3af8d46c8fa.zip | |
Add trailing slash when spidering URLs for versions (#8429)
By default, if a package does not specify a list_url and does not download from
a common repository, Spack runs dirname on the package URL. Given a URL
like https://root.cern.ch/download/root_v6.09.02.source.tar.gz, this returns
https://root.cern.ch/download. However, https://root.cern.ch/download
gives a 404, while https://root.cern.ch/download/ works just fine.
Note that some servers *don't* work with a trailing slash, so this tries with and
without the slash. This will double the number of URLs searched but the
slowdown should only affect the "spack versions" command.
| -rw-r--r-- | lib/spack/spack/util/web.py | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/lib/spack/spack/util/web.py b/lib/spack/spack/util/web.py index f694b49e7a..b9f94fcc35 100644 --- a/lib/spack/spack/util/web.py +++ b/lib/spack/spack/util/web.py @@ -286,6 +286,13 @@ def find_versions_of_archive(archive_urls, list_url=None, list_depth=0): for aurl in archive_urls: list_urls.add(spack.url.find_list_url(aurl)) + # Add '/' to the end of the URL. Some web servers require this. + additional_list_urls = set() + for lurl in list_urls: + if not lurl.endswith('/'): + additional_list_urls.add(lurl + '/') + list_urls.update(additional_list_urls) + # Grab some web pages to scrape. pages = {} links = set() |