woodpecker/pipeline/frontend/yaml/compiler/convert.go
2023-12-23 00:42:30 +01:00

253 lines
7.3 KiB
Go

// Copyright 2023 Woodpecker Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package compiler
import (
"fmt"
"maps"
"path"
"strings"
"github.com/oklog/ulid/v2"
backend_types "go.woodpecker-ci.org/woodpecker/v2/pipeline/backend/types"
"go.woodpecker-ci.org/woodpecker/v2/pipeline/frontend/metadata"
"go.woodpecker-ci.org/woodpecker/v2/pipeline/frontend/yaml/compiler/settings"
yaml_types "go.woodpecker-ci.org/woodpecker/v2/pipeline/frontend/yaml/types"
"go.woodpecker-ci.org/woodpecker/v2/pipeline/frontend/yaml/utils"
)
func (c *Compiler) createProcess(name string, container *yaml_types.Container, stepType backend_types.StepType) (*backend_types.Step, error) {
var (
uuid = ulid.Make()
detached bool
workingdir string
workspace = fmt.Sprintf("%s_default:%s", c.prefix, c.base)
privileged = container.Privileged
networkMode = container.NetworkMode
ipcMode = container.IpcMode
// network = container.Network
)
networks := []backend_types.Conn{
{
Name: fmt.Sprintf("%s_default", c.prefix),
Aliases: []string{container.Name},
},
}
for _, network := range c.networks {
networks = append(networks, backend_types.Conn{
Name: network,
})
}
extraHosts := make([]backend_types.HostAlias, len(container.ExtraHosts))
for i, extraHost := range container.ExtraHosts {
name, ip, ok := strings.Cut(extraHost, ":")
if !ok {
return nil, &ErrExtraHostFormat{host: extraHost}
}
extraHosts[i].Name = name
extraHosts[i].IP = ip
}
var volumes []string
if !c.local {
volumes = append(volumes, workspace)
}
volumes = append(volumes, c.volumes...)
for _, volume := range container.Volumes.Volumes {
volumes = append(volumes, volume.String())
}
// append default environment variables
environment := map[string]string{}
maps.Copy(environment, container.Environment)
maps.Copy(environment, c.env)
environment["CI_WORKSPACE"] = path.Join(c.base, c.path)
environment["CI_STEP_NAME"] = name
if stepType == backend_types.StepTypeService || container.Detached {
detached = true
}
if !detached || len(container.Commands) != 0 {
workingdir = c.stepWorkdir(container)
}
if !detached {
pluginSecrets := secretMap{}
for name, secret := range c.secrets {
if secret.Available(container) {
pluginSecrets[name] = secret
}
}
if err := settings.ParamsToEnv(container.Settings, environment, pluginSecrets.toStringMap()); err != nil {
return nil, err
}
}
if utils.MatchImage(container.Image, c.escalated...) && container.IsPlugin() {
privileged = true
}
authConfig := backend_types.Auth{}
for _, registry := range c.registries {
if utils.MatchHostname(container.Image, registry.Hostname) {
authConfig.Username = registry.Username
authConfig.Password = registry.Password
authConfig.Email = registry.Email
break
}
}
for _, requested := range container.Secrets.Secrets {
secret, ok := c.secrets[strings.ToLower(requested.Source)]
if ok && secret.Available(container) {
environment[strings.ToUpper(requested.Target)] = secret.Value
} else {
return nil, fmt.Errorf("secret %q not found or not allowed to be used", requested.Source)
}
}
// Advanced backend settings
backendOptions := backend_types.BackendOptions{
Kubernetes: convertKubernetesBackendOptions(&container.BackendOptions.Kubernetes),
}
memSwapLimit := int64(container.MemSwapLimit)
if c.reslimit.MemSwapLimit != 0 {
memSwapLimit = c.reslimit.MemSwapLimit
}
memLimit := int64(container.MemLimit)
if c.reslimit.MemLimit != 0 {
memLimit = c.reslimit.MemLimit
}
shmSize := int64(container.ShmSize)
if c.reslimit.ShmSize != 0 {
shmSize = c.reslimit.ShmSize
}
cpuQuota := int64(container.CPUQuota)
if c.reslimit.CPUQuota != 0 {
cpuQuota = c.reslimit.CPUQuota
}
cpuShares := int64(container.CPUShares)
if c.reslimit.CPUShares != 0 {
cpuShares = c.reslimit.CPUShares
}
cpuSet := container.CPUSet
if c.reslimit.CPUSet != "" {
cpuSet = c.reslimit.CPUSet
}
var ports []uint16
for _, port := range container.Ports {
ports = append(ports, uint16(port))
}
// at least one constraint contain status success, or all constraints have no status set
onSuccess := container.When.IncludesStatusSuccess()
// at least one constraint must include the status failure.
onFailure := container.When.IncludesStatusFailure()
failure := container.Failure
if container.Failure == "" {
failure = metadata.FailureFail
}
return &backend_types.Step{
Name: name,
UUID: uuid.String(),
Type: stepType,
Alias: container.Name,
Image: container.Image,
Pull: container.Pull,
Detached: detached,
Privileged: privileged,
WorkingDir: workingdir,
Environment: environment,
Commands: container.Commands,
ExtraHosts: extraHosts,
Volumes: volumes,
Tmpfs: container.Tmpfs,
Devices: container.Devices,
Networks: networks,
DNS: container.DNS,
DNSSearch: container.DNSSearch,
MemSwapLimit: memSwapLimit,
MemLimit: memLimit,
ShmSize: shmSize,
Sysctls: container.Sysctls,
CPUQuota: cpuQuota,
CPUShares: cpuShares,
CPUSet: cpuSet,
AuthConfig: authConfig,
OnSuccess: onSuccess,
OnFailure: onFailure,
Failure: failure,
NetworkMode: networkMode,
IpcMode: ipcMode,
Ports: ports,
BackendOptions: backendOptions,
}, nil
}
func (c *Compiler) stepWorkdir(container *yaml_types.Container) string {
if path.IsAbs(container.Directory) {
return container.Directory
}
return path.Join(c.base, c.path, container.Directory)
}
func convertKubernetesBackendOptions(kubeOpt *yaml_types.KubernetesBackendOptions) backend_types.KubernetesBackendOptions {
resources := backend_types.Resources{
Limits: kubeOpt.Resources.Limits,
Requests: kubeOpt.Resources.Requests,
}
var tolerations []backend_types.Toleration
for _, t := range kubeOpt.Tolerations {
tolerations = append(tolerations, backend_types.Toleration{
Key: t.Key,
Operator: backend_types.TolerationOperator(t.Operator),
Value: t.Value,
Effect: backend_types.TaintEffect(t.Effect),
TolerationSeconds: t.TolerationSeconds,
})
}
var securityContext *backend_types.SecurityContext
if kubeOpt.SecurityContext != nil {
securityContext = &backend_types.SecurityContext{
Privileged: kubeOpt.SecurityContext.Privileged,
RunAsNonRoot: kubeOpt.SecurityContext.RunAsNonRoot,
RunAsUser: kubeOpt.SecurityContext.RunAsUser,
RunAsGroup: kubeOpt.SecurityContext.RunAsGroup,
FSGroup: kubeOpt.SecurityContext.FSGroup,
}
}
return backend_types.KubernetesBackendOptions{
Resources: resources,
ServiceAccountName: kubeOpt.ServiceAccountName,
NodeSelector: kubeOpt.NodeSelector,
Tolerations: tolerations,
SecurityContext: securityContext,
}
}