Fix artifact registration in Releases API for Teleport Connect #13946
Conversation
| @@ -375,7 +375,7 @@ for file in $(find . -type f ! -iname '*.sha256' ! -iname '*-unsigned.zip*'); do | |||
| cat $WORKSPACE_DIR/curl_out.txt | |||
| exit 1 | |||
| fi | |||
| curl $CREDENTIALS --fail -o /dev/null -X PUT "$RELEASES_HOST/releases/$product@$VERSION/assets/$(basename $file)" | |||
| curl $CREDENTIALS --fail -o /dev/null -X PUT "$RELEASES_HOST/releases/$product@$VERSION/assets/$(basename "$file" | sed 's/ /%%20/g')" | |||
There was a problem hiding this comment.
Curl encodes data (body or url query), but not URLs, it expects them pre-encoded.
justinas
force-pushed
the
justinas/rlz-teleterm
branch
from
d8eeb13 to
077fd72
Compare
justinas
force-pushed
the
justinas/rlz-teleterm
branch
from
077fd72 to
f09a00c
Compare
| @@ -1228,7 +1231,7 @@ steps: | |||
| cat $WORKSPACE_DIR/curl_out.txt | |||
| exit 1 | |||
| fi | |||
| curl $CREDENTIALS --fail -o /dev/null -X PUT "$RELEASES_HOST/releases/$product@$VERSION/assets/$(basename $file)" | |||
| curl $CREDENTIALS --fail -o /dev/null -X PUT "$RELEASES_HOST/releases/$product@$VERSION/assets/$(basename "$file" | sed 's/ /%20/g')" | |||
There was a problem hiding this comment.
Is there a more general way to urlencode the filename?
There was a problem hiding this comment.
I couldn't find an easily available standalone utility to urlencode, aside from snippets like this https://stackoverflow.com/a/10660730/1264974
IMO it's fine for now, as our filenames are pretty uniform, and Teleport Connect is the first thing that needs to be urlencoded (all others roughly fall into the [A-z0-9._-] set)
| @@ -1206,20 +1206,23 @@ steps: | |||
| - which curl || apk add --no-cache curl | |||
| - |- | |||
| cd "$WORKSPACE_DIR/go/artifacts" | |||
| for file in $(find . -type f ! -iname '*.sha256' ! -iname '*-unsigned.zip*'); do | |||
| find . -type f ! -iname '*.sha256' ! -iname '*-unsigned.zip*' | while read -r file; do | |||
There was a problem hiding this comment.
What's the difference between for file in ... vs find ... | while read ... - asking for a bash deficient person (i.e. me)
There was a problem hiding this comment.
Not an expert either, but if I remember correctly: for will split by IFS, which by default includes spaces. One could in theory use something like IFS=$'\n' for ... to only split elements by newline, but I think this did not work as easily with alpine's shell (which does not have some of bash's stuff).
read X Y Z will read a line, and still split it to words (by spaces), but if you have more output variables than words, the remainder will end up in the last variable. E.g. with input foo bar baz quux, read X Y Z will end up in Z="baz quux". Thus, with only one output variable (read X), no splitting is performed, and in effect the whole line ends up in X unsplitted.
…13946) Fix artifact registration for Teleport Connect
…13946) Fix artifact registration for Teleport Connect
…13946) Fix artifact registration for Teleport Connect
…#13946) Fix artifact registration for Teleport Connect
Teleport Connect-X.Y.Z.dmghas recently been added to CI builds. This change fixes a couple of things to support it:teleport(OSS) andteleport-ent(Enterprise) releases - same logic as fortshpackages.This is undoubtedly somewhat hacky, but I hope to re-code this logic in a nicer way with the move to Go (https://github.com/gravitational/cloud/issues/1504). This change only concerns master and branch/v10, but I will try to port to older branches as well for consistency (also minimizes merge conflicts down the line).
All Dronegen changes auto-generated.