local checksum not match remote while node-gyp install #693
Comments
|
We're seeing the same issue -- see travis-ci/travis-ci#5562 |
|
What's the issue? It downloads the tarball and the checksum fails? What makes you think it's node-gyp and not e.g. a network problem? |
|
Quoting @klemmster in that issue:
|
|
That it's the correct file size doesn't mean they're the right bytes; chunks may have gotten corrupted in transit. What does I'm inclined to believe it's an issue local to your environment. If everyone was hit by this, we'd be swamped in bug reports, but we're not. |
for i in {1..100}; do curl -s "https://nodejs.org/dist/v0.10.36/node-v0.10.36.tar.gz" | sha256sum; doneprinted the correct sha256sum a hundred times Also, fs.readFile('/tmp/node-v0.10.36.tar.gz', function(err, data) {
if(err) {throw err;}
var shasum = crypto.createHash('sha256');
shasum.update(data);
console.log(shasum.digest('hex'));
});did. I wasn't able to find the tarball download by node-gyp. If you could point me to it, I'd be happy to provide a checksum for that as long as travis debugging instance is still available to me (don't know when they close it) |
|
v0.10.36 is the tarball you want? What happens when you |
|
|
And the same with --verbose |
|
Thanks. Can you try downloading v0.10.36 again but now with node v4.2.6 or v5.x as the binary that runs node-gyp? Please delete the .node-gyp directory first. |
|
Also, FWIW, it works perfectly for me with v0.10.41 and v6.0.0-pre (i.e. node master.) |
|
|
The "unexpected end of file" error means something went wrong while downloading. It's probably either a corrupted chunk or the connection was closed prematurely. |
|
Running that little script: var http = require('http');
var fs = require('fs');
var file = fs.createWriteStream("node-v0.10.36.tar.gz1");
var request = http.get("http://nodejs.org/dist/v0.10.36/node-v0.10.36.tar.gz", function(response) {
response.pipe(file);
});a 100 times: while it still never fails with curl. |
|
Your guess is as good as mine. I ran it in batches of 10 downloads a few times but I always get the right checksum: I did notice that node v0.10.40 uses a ton more CPU time than v6.0.0-pre but that doesn't seem to affect the download results: My working hypothesis is still that it's an issue local to your environment. |
|
On a newly-configured Windows 10 box , The correct size is 14,410,899 (running download.js on Linux shows this). So it looks like the last block is being arbitrarily lost. Whatever the cause, it makes node-gyp unusable on Windows. |
|
/cc @nodejs/platform-windows - is anyone able to reproduce? |
|
FWIW it is not a node-gyp issue, pretty sure of that. I also tried node-gyp-install and got the same results. Using 'node-gyp configure' I was after about 8 attempts able to get an OK, and then build. My test server is on Rackspace and has no proxy. However it does break node-gyp badly so is worth investigating IMO. |
|
@hintjens (and others able to repro this) can you Cruising through the |
|
node-gyp is using: |
|
well then .. that's the latest, I'll file an issue over there and see if they have a clue, we can't rule out a v0.10 problem, however. |
|
actually, I'm going to hold off and try a bit harder to reproduce it myself, there's some old style streams work going on in there that I think could be improved |
|
I never have this problem on Linux or OS X, almost always have this problem with Windows (now that I'm doing prebuilds for all platforms of native dependencies). |
|
Fwiw, in eight Travis builds we are seeing five checksum mismatches, one "Completion callback never invoked!", one 522, and one success, suggesting that this is network-related. |
|
we are having issues with our download service right now. hoping to have it
sorted out very soon
…On Fri, Dec 9, 2016, 11:36 AM Chad Whitacre ***@***.***> wrote:
Fwiw, in eight Travis builds
<gratipay/gratipay.com#4130 (comment)>
we are seeing five checksum mismatches, one "Completion callback never
invoked!", one 522, and one success, suggesting that this is
network-related.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#693 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAecV8U3FHYNyv00TdvYqu-MT9epVDbVks5rGYOUgaJpZM4FtWv0>
.
|
|
closing due to staleness |
#645
The text was updated successfully, but these errors were encountered: