From 6da17f59305c7829e740b870d89eb09ef3ddf41e Mon Sep 17 00:00:00 2001
From: Kyle McGill <kmcgill@nvidia.com>
Date: Tue, 21 Jan 2025 13:07:38 -0800
Subject: [PATCH] Downgrade patchelf version from 0.18.0 to 0.17.2 due to
 patchelf regression

    Patchelf shipped a regression in 0.18.0 and has since yanked the pypi release pointing to
    0.17.2 as the most recent version. However, 0.18.0 is still the version shipped in both the apt and yum
    repositories, thus we must use pip to install the version we want.
    See https://github.com/mayeut/patchelf-pypi/issues/87
---
 README.md | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 8ed211f..2b27283 100644
--- a/README.md
+++ b/README.md
@@ -52,7 +52,8 @@ main Triton [issues page](https://github.com/triton-inference-server/server/issu
 Use a recent cmake to build. First install the required dependencies.
 
 ```
-$ apt-get install patchelf rapidjson-dev python3-dev
+$ apt-get install rapidjson-dev python3-dev python3-pip
+$ pip3 install patchelf==0.17.2
 ```
 
 An appropriate PyTorch container from [NGC](https://ngc.nvidia.com) must be used.