URL without filename fails

[ryandesign]

With wcurl 2024-07-02:

% wcurl https://github.com/Debian/     
curl: Remote file name has no length
curl: (23) Failed writing received data to disk/application

However with wget 1.24.5:

% wget https://github.com/Debian/
--2024-07-04 09:53:47--  https://github.com/Debian/
Resolving github.com (github.com)... 140.82.114.3
Connecting to github.com (github.com)|140.82.114.3|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: unspecified [text/html]
Saving to: ‘index.html’

index.html                                                             [  <=>                                                                                                                                                          ] 245.03K  1.17MB/s    in 0.2s    

2024-07-04 09:53:48 (1.17 MB/s) - ‘index.html’ saved [250913]

[bagder]

There's a plan to fix this in curl, although saving it with a different filename than what wget picks: curl/curl#13988

[ryandesign]

This works for me but please test:

https://salsa.debian.org/debian/wcurl/-/merge_requests/4

Note there is a new dependency on trurl.

[samueloph]

I'll try to keep the discussion about this issue on salsa, but if anyone would like to reply and doesn't have an account, feel free to do it here.

[BrianInglis]

Works for me but saves file Debian not Debian.html.
With just a host name, e.g. curl.se saves curl_response, not even curl-response.html, or better curl[-.]se.html or curl[-.]se[-.]index.html, which would be better than wget/2 anonymous index.html! Added similar comment to @curl #13988
Just packaged wcurl as part of Cygwin distribution standard main package curl 8.10 so trying to get ahead of users trying it out!
I describe wcurl and mention your home page in the announcement, so they could come here ;^>
No other Cygwin packagers had any comments on whether I should include it in curl, make it a subpackage of curl source package, or package wcurl source and "binary" separately, so thought I would help out most users by giving out a free wcurl script and docs with every curl command line package. ;^>

[BrianInglis]

Could translate back from response content-type: header media-type/mime-type, for example:

$ curl -I curl.se
...
HTTP/2 200
server: nginx/1.21.1
content-type: text/html
...

to file type suffix extension using shared-mime-info data in /usr/share/mime/packages/freedesktop.org.xml which gives a list of glob patterns for each mime-type, for example:

$ awk '/<mime-type\stype="text\/html">/,/<\/mime-type>/' /usr/share/mime/packages/freedesktop.org.xml
  <mime-type type="text/html">
    <comment>HTML document</comment>
    <comment xml:lang="zh_TW">HTML 文件</comment>
    <comment xml:lang="zh_CN">HTML 文档</comment>
...
    <comment xml:lang="en_GB">HTML document</comment>
...
    <acronym>HTML</acronym>
    <expanded-acronym>HyperText Markup Language</expanded-acronym>
    <sub-class-of type="text/plain"/>
    <magic>
      <match type="string" value="&lt;!DOCTYPE HTML" offset="0:256"/>
...
    </magic>
    <magic priority="40">
      <match type="string" value="&lt;!--" offset="0"/>
      <match type="string" value="&lt;TITLE" offset="0:256"/>
      <match type="string" value="&lt;title" offset="0:256"/>
    </magic>
    <glob pattern="*.html" weight="80"/>
    <glob pattern="*.htm" weight="80"/>
  </mime-type>

The code could be something equivalent to this awk command:

$ awk '/<mime-type\s+type="[^"]+"[^>]*>/,/<\/mime-type>/ {
  if (!found) found = match( $0, "<mime-type type=\"" mime_type "\"");
  if (found && /<glob\s+pattern="/) {
    sub( /^\s*<glob\s+pattern="\*/, "");
    sub( /".*$/, "");
    print;
    exit; # exit on first match
  }
}' mime_type="text/html" /usr/share/mime/packages/freedesktop.org.xml
.html