[sending this email again as I forgot to reply to the list]
On 3/21/19 11:00 PM, Aaron Lawrence wrote:
Hi all,
I was running TortoiseHg 4.8.2. On our largest repo (16,000 commits,
6GB) I started getting a HTTP 400 error when pulling.
unrelated-remark: 6GB for 16,000 commit is fairly large. Do you have a
lot of large file committed ? Or do you have a very branchy (many merge)
repository ? What are the sizes of the .hg/store/00manifest.d` and
`.hg/store/00changelog.d` files?
Fiddler proxy showed HTTP Error 400. The size of the request headers is
too long.
Can you share a trace with of a pull using the `--debug` flag ? And one
using both `--debug` and `--config devel.debug.peer-request=yes`
I noticed that going back to TortoiseHg 3.7.3 actually stopped the
problem happening.
Now I’m running with 3.7.3 and it seems fine. Other people are also
using older Hg with the same repo and not having any problem.
We don’t want to be stuck on an old version of Mercurial though.
I found people talking about having too many dangling heads. Our repo
has 200 topological heads, which seems well below what others had
problems with. But more importantly, when I got rid of 50 of them my
merging to a graveyard branch, it didn’t help, the 400 error kept
happening with 4.8.2.
200 heads is well below the problematic range. Are you using named
branches ? bookmarks ? If so how many of each ?
If I did a fresh clone, it would work for a little while then start
giving 400 errors again.
That's is odd.
I could see that the headers went from ~8k to 26k without anything changing.
Can you share some extra details on the http request with such header
inflation ?
So I don’t know what to think – why would older mercurial work better?
Why does the header request size increase after a few pulls?
Our server is an old version of Kiln, 2.9.61.0 (DB 202) running on IIS.
That looks ancient. Have you though about upgrading to a newer version
for your Mercurial server?
--
Pierre-Yves David
_______________________________________________
Mercurial mailing list
[email protected]
https://www.mercurial-scm.org/mailman/listinfo/mercurial
